skills/tristanmanchester__agent-skills
1
0
Fork 0
mirror of https://github.com/tristanmanchester/agent-skills.git synced 2026-09-14 17:42:18 +08:00
Code Issues Packages Projects Releases Wiki Activity

Publish current skills catalogue and verified merge-review records (#25)

Browse Source 
* Document complete skills catalogue and verified modernisation checkpoint

* Complete catalogue dispositions for all 37 reviewed skills

* Record final review dispositions, new skill PRs, and explicit validation limits

* Record completed review coverage, retained skills, and final batch verification

* Record recovered overhaul state and targeted template verification

* Advance modernisation checkpoint with verified Skia and Textual follow-up heads

* Collect a read-only review checkout and PR evidence for integration validation
This commit is contained in:
Tristan Manchester
2026-09-14 10:18:43 +02:00
committed by GitHub
parent 2572ed26fe
commit 3323bc919a
5 changed files with 494 additions and 15 deletions
Download Patch File Download Diff File Expand all files Collapse all files
MAINTENANCE.md
+144
View File
@@ -0,0 +1,144 @@
# Skills modernisation review and publication record
Snapshot: September 13, 2026. Source baseline:
`b28669d18b1731233a07fb73da3d00b39d41bee2`.
**All 37 baseline skill directories have a source-review disposition.** There
are 34 open skill PRs, one closed-unmerged skill PR, and two skills retained
without changes. The final continuation published 13 skill PRs (#26–#38), keeping
one skill per PR. Repository-only catalogue work remains in #25. No PR was merged
by this work; main was rechecked and still points to the baseline.
This completes the disposition pass, not exhaustive runtime certification of every
retained file or upstream fact. The individual PRs identify inspected source,
current primary-source checks, tests actually run, and important gaps. A source
review does not establish live service access, successful native builds, complete
accessibility, clinical efficacy, or improved agent output.
## Recover or review the work
Read the current default branch and actual open/closed PR state before acting on
this dated record. Reuse the existing skill branch/PR for further repairs rather
than creating a duplicate. A PR description reports work; it is not independent
proof of correctness. Read the changed code and relevant upstream contract.
**Merge #9 before #10:** retiring meta-ads-control depends on the capability-preserving
meta-ads-cli update. Other skill PRs were independently based on the same baseline,
not stacked with unrelated changes. Merging this catalogue does not merge those
skills, and an ordinary install from main does not include unmerged changes.
**#22 was observed closed without merging** (write-good-docs). It was not reopened,
recreated, or treated as accepted. Its prior packaging proposal remains available
in history; any further action needs the user's direction. Do not silently turn a
closed proposal into a new PR under a different name.
The machine-readable [checkpoint](maintenance/modernisation-progress.json) records
all 37 directories, their PR/disposition, latest published head commits, and test
counts. Reconcile it after review decisions or merges.
## Final publication batch
| PR | Skill | Verified repair or modernisation |
| --- | --- | --- |
| [#26](https://github.com/tristanmanchester/agent-skills/pull/26) | fabric-api | Official SDK first, explicit Memory API base scope, signed-upload and ambiguous-write recovery; inherited detailed routes labelled for current verification |
| [#27](https://github.com/tristanmanchester/agent-skills/pull/27) | fabric-cli | Native CLI, no false ask-read guarantee or blind JSON-failure replay; explicit resource type, workspace, and memory scope |
| [#28](https://github.com/tristanmanchester/agent-skills/pull/28) | animating-react-native-expo | Current RNGH3 cancellation, gesture offsets/pending state, compatible native stack, reduced motion and cleanup |
| [#29](https://github.com/tristanmanchester/agent-skills/pull/29) | react-native-skia | Current dependency-delivered binaries and animation contracts; remove obsolete postinstall/setup scoring while keeping drawing methods/templates |
| [#30](https://github.com/tristanmanchester/agent-skills/pull/30) | nextjs-framer-motion-animations | Framework route-transition versus local presence contracts, semantic controls and boundaries; remove heuristic scaffolding/import rewriting |
| [#31](https://github.com/tristanmanchester/agent-skills/pull/31) | textual-tui | Precise worker lifecycle plus bounded, non-writing starter previews and AST-safe titles; keep all eleven starter archetypes |
| [#32](https://github.com/tristanmanchester/agent-skills/pull/32) | reddit | Approved read-only access instead of anonymous-JSON retries; preserve original context, dates, truncation, and thread completeness |
| [#33](https://github.com/tristanmanchester/agent-skills/pull/33) | tracking-pettracer-location | Measurement-time freshness, finite coordinates, fixed-origin bounded reads, and raw unverified quality fields; no untested live watcher promise |
| [#34](https://github.com/tristanmanchester/agent-skills/pull/34) | wordly-wisdom | Fixed utility anchors avoid option-set rank reversal; strict finite probabilities/weights and explicit scenario/model assumptions |
| [#35](https://github.com/tristanmanchester/agent-skills/pull/35) | relationship-science-coach | Remove keyword-based safety decisions; preserve practical coaching with context, consent, autonomy, and calibrated evidence |
| [#36](https://github.com/tristanmanchester/agent-skills/pull/36) | ai-codebase-deep-modules | Correct import/visibility enforcement and remove unchecked placeholder generator; retain ownership/contracts/testing methods |
| [#37](https://github.com/tristanmanchester/agent-skills/pull/37) | good-services-service-design | Evidence states and scoped local scoring, distinct from the author's scale; proportional deliverables and preserved end-to-end service methods |
| [#38](https://github.com/tristanmanchester/agent-skills/pull/38) | designing-beautiful-websites | Distinctive, task-scoped design plus current selected accessibility checks; retain contrast helper and specialist references |
These are proposed changes, not claims that providers, stores, users, or reviewers
have accepted them. Removing a handwritten client in favour of a maintained
surface does not prove that surface covers every historical operation. The skill
must verify the specific required operation rather than quietly dropping it.
## Two skills retained without changes
### rust-anti-slop
Read the core, full rules reference, workspace-lints.toml, and clippy.toml. The
policy deliberately balances errors, unsafe code, cloning, concurrency, narrow
lint exceptions, and workspace inheritance. Selected lint availability and Cargo
configuration were checked against official documentation; no identified defect
justified changing the user's intentional Rust policy merely to make it look new.
This is a keep decision within that review scope, not a claim every Clippy option
was executed on every supported toolchain. No cargo/rustc/lint run was performed.
Selected sources: https://doc.rust-lang.org/clippy/ and
https://doc.rust-lang.org/cargo/reference/manifest.html#the-lints-section.
### giving-presentations
Read the core, source notes, and slide-design/visuals reference. The audience,
message, narrative, evidence, live-versus-async, rehearsal, quantitative integrity,
and readable/accessible visual methods remain useful. There is no API migration
or concrete packaging defect identified in those reviewed paths requiring a PR.
Timing, word-count, and design ranges remain heuristics, not universal empirical
laws. The whole inherited source bibliography was not freshly re-audited, and no
presentation, audience study, reading-order test, or model-output comparison was
run. Keeping it avoids a cosmetic rewrite of a useful method.
## Tests actually executed in this continuation
Thirty Python unittest methods passed on Linux / Python 3.13.5, and all three
suites passed again in the final check:
| Skill | Methods | What they establish |
| --- | ---: | --- |
| textual-tui | 8 | Literal quoted/injection-like titles, identifier validation, non-writing previews, complete preflight, exclusive output, symlink/path safeguards |
| tracking-pettracer-location | 10 | Measurement-time/coordinate semantics, explicit device selection, pre-network window validation, fixed origin, no retries/redirects, bounded mocked response |
| wordly-wisdom | 12 | Fixed-anchor invariance, finite/strict inputs, direction/bounds, EV arithmetic, duplicate/oversized JSON, CLI failures, input preservation |
The source files used locally were reconstructed selected files, not a full repo
clone. Network access from the execution container was unavailable; connected
GitHub performed the repository writes. Selected local implementation Git blob
hashes were compared with the published files, including the Textual scaffolder
and all three decision-math modules. PetTracer's script/test hashes are recorded
in its PR. No target account credentials or private user data were used in tests.
The Textual tests use synthetic templates; the full eleven-template self-check,
Textual/Pilot app runtime, and terminal/browser rendering were not run. PetTracer
transport is mocked; its unofficial live protocol, quality units, battery calibration,
and history completeness remain unverified. Decision arithmetic does not validate
the supplied utilities/probabilities or predict the user's outcome.
The other ten new skill PRs received source/documentation review rather than
executable regression suites. No authenticated Fabric/Reddit/PetTracer/provider
request, native build, store upload, browser/screen-reader test, paid job, clinical
study, or controlled agent-output evaluation was performed. Earlier batches' tests
were not all rerun; their reports remain scoped to their own PRs.
## Final cross-reference and status checks
Rechecked the decision skill's new core links against its actual tree and inspected
the retained scenario sample against the new input shape. Inspected current Skia
performance/evaluation references and indexed references to removed setup helpers;
these checks were targeted, not a complete link crawl of every retained file.
New web-design links were checked against its retrieved recursive tree.
The checkpoint was locally parsed/checked for 37 unique directories, 35 distinct
skill PR mappings, the 13 new PR numbers/head formats, two keep dispositions,
one closed-unmerged disposition, and the 30-method test sum. Live GitHub reported
35 open PRs total at final status check: 34 skill PRs plus repository PR25.
## Remaining verification before real use
Review and merge deliberately, then install the intended revision. Version-match
SDK/native types and run the actual integration paths that matter to the app or
account. Test live operations only within explicit authorisation and appropriate
test data. Preserve unresolved outcomes and never use a timeout as proof a mutation
did not occur. Evaluate skills on real representative agent tasks before claiming
better outputs. These are documented validation limits, not unassigned skill reviews.
Current-source entry points for this batch are linked in each skill PR. Apply
https://agentskills.io/specification and task-specific primary sources, preserving
provenance, licensing, and a clear separation between a method, a supplied fixture,
an observed test result, and a proposed improvement.
README.md
+87 -15
View File
@@ -1,26 +1,98 @@
# skills
# Agent skills
[![Quick install](https://img.shields.io/static/v1?label=quick%20install&message=npx%20skills%20add%20tristanmanchester%2Fagent-skills&color=4c1&logo=terminal&logoColor=white)](https://github.com/tristanmanchester/agent-skills)
[![skills.sh](https://img.shields.io/badge/skills.sh-directory-0b7285)](https://skills.sh/)
Task-focused instructions, references, templates, and tools for skills-compatible
agents. Each active skill has its own directory and `SKILL.md` entry point.
## Install and use
Collection of Clawdbot skills published on skills.sh. Each skill lives in its own folder with a SKILL.md prompt and supporting scripts/assets.
## Quick install
Read the selected skill's entry point, licence, dependencies, and account
requirements. The [Skills CLI](https://skills.sh/docs/cli) supports:
```bash
npx skills add tristanmanchester/agent-skills
```
## Skills
This runs a CLI and changes the selected agent's skill installation. Select the
skills relevant to your work, or use your host's supported installation method.
Keep each entry point with its relative supporting files and record the source
revision. Do not keep duplicate active copies of the same skill.
- assemblyai-transcribe: Transcribe audio/video with AssemblyAI (local upload or URL), plus subtitles and paragraph/sentence exports.
- designing-beautiful-websites: Designs UX/UI for websites and web apps (strategy, IA, flows, wireframes, systems).
- fabric-api: Create/search Fabric resources via HTTP API (notepads, folders, bookmarks, files).
- ios-simulator: Automate iOS Simulator workflows (simctl/idb, install/launch, screenshots, UI taps).
- reddit-readonly (folder `reddit`): Browse and search Reddit in read-only mode using public JSON endpoints.
- track17: Track parcels via the 17TRACK API (local SQLite DB, polling plus optional webhooks).
Start with `SKILL.md`; load the references, assets, or scripts it identifies when
needed. Not every skill has a separate README. Resolve bundled scripts from the
installed skill directory, not an unrelated project's `scripts/` directory.
Instructions do not install dependencies, establish service access, or authorise
live mutations. Licences apply per skill; there is no blanket repository licence.
## Usage
## Catalogue
See each skill folder's README.md and SKILL.md for details.
The repository contains **36 active skill directories**. The September 2026
overhaul reviewed 37 baseline directories: 33 update PRs and one retirement PR
have merged, two skills were retained unchanged, and the `write-good-docs`
packaging proposal (#22) remains closed unmerged. Its existing skill is retained.
`meta-ads-control` was retired after its useful operations were consolidated into
[meta-ads-cli](meta-ads-cli/SKILL.md). The required merge order, #9 before #10, was
verified. There is no replacement compatibility entry point.
| Skill / entry point | Purpose |
| --- | --- |
| [agent-use](agent-use/SKILL.md) | Design and audit agent-facing capabilities, contracts, and recovery. |
| [ai-codebase-deep-modules](ai-codebase-deep-modules/SKILL.md) | Design cohesive modules, explicit ownership, and enforceable boundaries. |
| [animating-react-native-expo](animating-react-native-expo/SKILL.md) | Build native animations and gestures with a compatible Expo stack. |
| [assemblyai-transcribe](assemblyai-transcribe/SKILL.md) | AssemblyAI transcription workflows and speaker-aware local exports. |
| [audit-openclaw-security](audit-openclaw-security/SKILL.md) | Review OpenClaw exposure, policy, secrets, and diagnostic handling. |
| [auditing-appstore-readiness](auditing-appstore-readiness/SKILL.md) | Assess source, archive, runtime, and store submission evidence separately. |
| [designing-beautiful-websites](designing-beautiful-websites/SKILL.md) | Design and review purposeful, distinctive, accessible websites. |
| [display-quantitative-information](display-quantitative-information/SKILL.md) | Create and critique faithful quantitative displays. |
| [exa-search](exa-search/SKILL.md) | Use explicitly selected Exa search and content-retrieval workflows. |
| [expo-revenuecat-superwall-integration](expo-revenuecat-superwall-integration/SKILL.md) | Integrate purchases, offers, identity, restores, and entitlements. |
| [extracting-mistral-ocr](extracting-mistral-ocr/SKILL.md) | Mistral document extraction with bounded, organised exports. |
| [fabric-api](fabric-api/SKILL.md) | Build Fabric.so SDK/API integrations and upload workflows. |
| [fabric-cli](fabric-cli/SKILL.md) | Operate Fabric.so through its native CLI and scoped project memory. |
| [generating-novel-ideas](generating-novel-ideas/SKILL.md) | Explore distinct mechanisms, critique concepts, and define useful tests. |
| [giving-presentations](giving-presentations/SKILL.md) | Develop audience-focused narratives, visuals, and rehearsal plans. |
| [good-services-service-design](good-services-service-design/SKILL.md) | Review whole services with explicit evidence and scoped scoring. |
| [ios-simulator](ios-simulator/SKILL.md) | Run and inspect simulator apps with explicit targets and checked outcomes. |
| [jax-development](jax-development/SKILL.md) | Develop, debug, profile, and benchmark JAX code. |
| [meta-ads-cli](meta-ads-cli/SKILL.md) | Use the official Meta Ads CLI with bounded advanced API operations. |
| [nature-article-writer](nature-article-writer/SKILL.md) | Prepare scientific manuscripts with journal-specific checks. |
| [nextjs-framer-motion-animations](nextjs-framer-motion-animations/SKILL.md) | Implement Motion interactions and appropriate Next.js transition boundaries. |
| [optimising-expo-react-native-performance](optimising-expo-react-native-performance/SKILL.md) | Measure and improve native performance against reproducible baselines. |
| [parallel-ai-search](parallel-ai-search/SKILL.md) | Use explicitly selected Parallel search, research, and bounded job workflows. |
| [react-native-skia](react-native-skia/SKILL.md) | Build custom native graphics, gestures, snapshots, and shaders. |
| [reddit](reddit/SKILL.md) | Read-only research through permitted web/connector access or an approved OAuth API client; no anonymous JSON client. |
| [relationship-science-coach](relationship-science-coach/SKILL.md) | Practical relationship coaching with contextual safety and calibrated evidence. |
| [resend-api](resend-api/SKILL.md) | Build Resend SDK/API integrations, verified events, and Automations. |
| [resend-cli](resend-cli/SKILL.md) | Operate Resend through its native CLI with deliberate send and campaign scope. |
| [rust-anti-slop](rust-anti-slop/SKILL.md) | Apply the repository's intentional Rust code and lint policy. |
| [styling-nativewind-v4-expo](styling-nativewind-v4-expo/SKILL.md) | Configure and use the explicitly versioned NativeWind v4 stack. |
| [textual-tui](textual-tui/SKILL.md) | Build and test Textual interfaces, workers, and starter applications. |
| [todoist-api](todoist-api/SKILL.md) | Use Todoist's official CLI/API with explicit bulk-write recovery. |
| [track17](track17/SKILL.md) | Track parcels and ingest authenticated 17TRACK updates. |
| [tracking-pettracer-location](tracking-pettracer-location/SKILL.md) | Inspect owner-authorised PetTracer fixes and history through a bounded unofficial reader. |
| [wordly-wisdom](wordly-wisdom/SKILL.md) | Analyse decisions with explicit assumptions and checked arithmetic. |
| [write-good-docs](write-good-docs/SKILL.md) | Write and edit evidence-preserving technical documentation; packaging proposal #22 remains closed unmerged. |
## Review records
The [September 14 PR review](maintenance/pr-review-20260914.md) records the final
skill merges, checked bot findings, local tests, and remaining validation limits.
The [machine-readable checkpoint](maintenance/modernisation-progress.json) records
all baseline dispositions. GitHub remains the authority for later PR state.
The [September 13 source review](MAINTENANCE.md) and
[early September 14 template follow-up](maintenance/follow-up-20260914.md) are
historical records. Their open-PR counts describe those earlier checkpoints,
not the current repository. Merging source does not update existing installations
or establish live-provider, native-device, or agent-output performance.
## Contributing
Follow the [Agent Skills format](https://agentskills.io/specification): precise
activation, task-focused instructions, and supporting material loaded when useful.
Preserve useful methods and verified helpers. Check changing contracts against
current primary sources, and record which tests actually ran.
Keep one skill's changes in one PR. Repository-wide catalogue and maintenance
changes belong in a separate PR. A fixture, syntax check, mocked test, or merged
PR is not evidence that a live integration or a representative agent task passed.
maintenance/follow-up-20260914.md
+85
View File
@@ -0,0 +1,85 @@
# Follow-up verification — September 14, 2026
Recovery found the remaining thirteen skill PRs (#26–#38) already published by
the interrupted run. The current catalogue has a disposition for every one of
the 37 baseline skills: 34 open skill PRs, one closed-unmerged skill PR (#22), and
two retained skills. Repository-only PR25 is additional. No new PR or merge was
made in this follow-up. Main remains b28669d18b1731233a07fb73da3d00b39d41bee2.
This is completion of source-review dispositions and proposed changes, not
exhaustive execution of every helper, native integration, or inherited reference.
The [September 13 record](../MAINTENANCE.md) describes the preceding batch. This
follow-up inspected retained examples, not just the updated skill entry points.
## Existing PR29 — React Native Skia
Latest head: `dae6239b4a0b938e40d36b7cd5275b98a52a922a`.
The retained pan/zoom template still used older Gesture builders, nested shared
values in a transform array, unbounded translation, zoom without focal-point
preservation, and intrinsic-pixel/density sizing rather than viewport fitting.
The replacement and its pure helper use RNGH3, one derived numeric transform,
viewport-centred coordinates, fitted aspect ratio, clamped ratios/bounds, and
one owner for two-finger translation. Native controls provide zoom/move/reset;
source/dimension changes reset view state. Integration notes identify both files
to copy and the remaining native acceptance checks.
The snapshot template also now serialises capture, waits for nonzero layout,
checks layout revision/unmount before accepting completion, and reports errors.
The static text card does not pretend layout alone establishes asynchronous
font/image/video readiness for an adapted app.
Executed on Linux, Node22.16.0 and TypeScript5.8.3:
- Strict compilation of the pure geometry module and **12 passing Node tests**.
- Tests cover aspect ratio, successive zooms, focal invariance, clamped ratios,
finite input/bounds, immutable state, and a deterministic 500-update sequence.
- Syntax transpilation of both changed TSX components without syntax errors.
No installed native-package type-check, Worklets transformation, Skia rendering,
gesture arbitration, device, browser, or screen-reader test ran. Snapshot runtime
was not tested. The geometry tests are not proof of native interaction behaviour.
## Existing PR31 — Textual
Latest head: `f8149ebbec737718c662ed2ef7d5a0b786199919`.
The retained chat starter could cancel an accepted prompt when another arrived,
re-anchored history on every append, interpolated prompts into Markdown, and
used a fixed-delay test which checked only message count. The real template now
preserves pending prompts/drafts, has explicit Stop behaviour and request revisions,
ignores stale results/events, handles visible failure state, and renders literal
text. Conditional follow behaviour rechecks whether the reader moved before the
refresh callback. Cancellation revokes local ownership before asking the worker
to stop; it makes no claim that a remote operation was undone.
**12 Python unittest methods passed** on Python3.13.5 using the real rendered
chat template's methods with explicit UI/worker doubles. Both real Python chat
templates parsed after placeholder replacement. Three generated Pilot tests were
added using controlled events and bounded completion rather than fixed sleeps.
Textual is not installed here. Pilot, TCSS, real worker scheduling, terminal/browser
rendering, and all eleven starter archetypes were not executed. These method tests
are distinct from the preceding batch's eight synthetic scaffold-engine tests,
which were not rerun in this follow-up. Repeated test runs are not additional tests.
## Publication verification
Selected published Git blob IDs match the local files used for these checks:
| File | Git blob |
| --- | --- |
| Skia pan/zoom math | `497458a716258f1a2225f9fb0a7c6a8e3c67ff45` |
| Skia snapshot template | `382110a18be5e184ca5858013784d594b5167a85` |
| Textual chat app template | `cc30eacd299cec1f6dad754274e57d5de2d0c498` |
| Textual method tests | `1dcaadd9d584b3a1d5177908ddd2c2da367882d1` |
GitHub returned 35 open PRs total: 34 skill proposals plus PR25. PR22 was rechecked
closed and unmerged; it was not reopened or recreated. Keep the #9-before-#10
merge dependency. Updating this record does not install or merge the skills.
The follow-up adds **24 distinct passing local tests**. It does not re-run or
certify the earlier batches. Current first-party contracts checked are linked in
`react-native-skia/references/pan-zoom-template.md`, the Skia snapshot PR comment,
and `textual-tui/references/chat-template.md`. No paid/provider/account action or
private user data was used for verification.
maintenance/modernisation-progress.json
+81
View File
@@ -0,0 +1,81 @@
{
"schema_version": 3,
"snapshot_date": "2026-09-14",
"scope": "Baseline skill dispositions after the last skill PR merged; repository housekeeping is published separately through PR25.",
"baseline_commit": "b28669d18b1731233a07fb73da3d00b39d41bee2",
"main_at_skill_merge_completion": "2572ed26fee46da651431bb7a4aaf2b3c5a9ed5f",
"counts": {
"baseline_skills": 37,
"active_skill_directories": 36,
"merged_update_prs": 33,
"merged_retirement_prs": 1,
"open_skill_prs": 0,
"closed_unmerged_skill_prs": 1,
"retained_without_change": 2,
"source_reviews_outstanding": 0
},
"skills": [
{"directory":"agent-use","status":"merged","present_in_main":true,"pr":21,"pr_url":"https://github.com/tristanmanchester/agent-skills/pull/21"},
{"directory":"ai-codebase-deep-modules","status":"merged","present_in_main":true,"pr":36,"pr_url":"https://github.com/tristanmanchester/agent-skills/pull/36"},
{"directory":"animating-react-native-expo","status":"merged","present_in_main":true,"pr":28,"pr_url":"https://github.com/tristanmanchester/agent-skills/pull/28"},
{"directory":"assemblyai-transcribe","status":"merged","present_in_main":true,"pr":18,"pr_url":"https://github.com/tristanmanchester/agent-skills/pull/18"},
{"directory":"audit-openclaw-security","status":"merged","present_in_main":true,"pr":3,"pr_url":"https://github.com/tristanmanchester/agent-skills/pull/3"},
{"directory":"auditing-appstore-readiness","status":"merged","present_in_main":true,"pr":17,"pr_url":"https://github.com/tristanmanchester/agent-skills/pull/17"},
{"directory":"designing-beautiful-websites","status":"merged","present_in_main":true,"pr":38,"pr_url":"https://github.com/tristanmanchester/agent-skills/pull/38"},
{"directory":"display-quantitative-information","status":"merged","present_in_main":true,"pr":23,"pr_url":"https://github.com/tristanmanchester/agent-skills/pull/23"},
{"directory":"exa-search","status":"merged","present_in_main":true,"pr":15,"pr_url":"https://github.com/tristanmanchester/agent-skills/pull/15"},
{"directory":"expo-revenuecat-superwall-integration","status":"merged","present_in_main":true,"pr":8,"pr_url":"https://github.com/tristanmanchester/agent-skills/pull/8"},
{"directory":"extracting-mistral-ocr","status":"merged","present_in_main":true,"pr":5,"pr_url":"https://github.com/tristanmanchester/agent-skills/pull/5"},
{"directory":"fabric-api","status":"merged","present_in_main":true,"pr":26,"pr_url":"https://github.com/tristanmanchester/agent-skills/pull/26"},
{"directory":"fabric-cli","status":"merged","present_in_main":true,"pr":27,"pr_url":"https://github.com/tristanmanchester/agent-skills/pull/27"},
{"directory":"generating-novel-ideas","status":"merged","present_in_main":true,"pr":24,"pr_url":"https://github.com/tristanmanchester/agent-skills/pull/24"},
{"directory":"giving-presentations","status":"retained_after_source_review","present_in_main":true},
{"directory":"good-services-service-design","status":"merged","present_in_main":true,"pr":37,"pr_url":"https://github.com/tristanmanchester/agent-skills/pull/37"},
{"directory":"ios-simulator","status":"merged","present_in_main":true,"pr":12,"pr_url":"https://github.com/tristanmanchester/agent-skills/pull/12"},
{"directory":"jax-development","status":"merged","present_in_main":true,"pr":19,"pr_url":"https://github.com/tristanmanchester/agent-skills/pull/19"},
{"directory":"meta-ads-cli","status":"merged","present_in_main":true,"pr":9,"pr_url":"https://github.com/tristanmanchester/agent-skills/pull/9"},
{"directory":"meta-ads-control","status":"retired_merged","present_in_main":false,"pr":10,"pr_url":"https://github.com/tristanmanchester/agent-skills/pull/10"},
{"directory":"nature-article-writer","status":"merged","present_in_main":true,"pr":20,"pr_url":"https://github.com/tristanmanchester/agent-skills/pull/20"},
{"directory":"nextjs-framer-motion-animations","status":"merged","present_in_main":true,"pr":30,"pr_url":"https://github.com/tristanmanchester/agent-skills/pull/30"},
{"directory":"optimising-expo-react-native-performance","status":"merged","present_in_main":true,"pr":6,"pr_url":"https://github.com/tristanmanchester/agent-skills/pull/6"},
{"directory":"parallel-ai-search","status":"merged","present_in_main":true,"pr":16,"pr_url":"https://github.com/tristanmanchester/agent-skills/pull/16"},
{"directory":"react-native-skia","status":"merged","present_in_main":true,"pr":29,"pr_url":"https://github.com/tristanmanchester/agent-skills/pull/29"},
{"directory":"reddit","status":"merged","present_in_main":true,"pr":32,"pr_url":"https://github.com/tristanmanchester/agent-skills/pull/32"},
{"directory":"relationship-science-coach","status":"merged","present_in_main":true,"pr":35,"pr_url":"https://github.com/tristanmanchester/agent-skills/pull/35"},
{"directory":"resend-api","status":"merged","present_in_main":true,"pr":14,"pr_url":"https://github.com/tristanmanchester/agent-skills/pull/14"},
{"directory":"resend-cli","status":"merged","present_in_main":true,"pr":13,"pr_url":"https://github.com/tristanmanchester/agent-skills/pull/13"},
{"directory":"rust-anti-slop","status":"retained_after_source_review","present_in_main":true},
{"directory":"styling-nativewind-v4-expo","status":"merged","present_in_main":true,"pr":7,"pr_url":"https://github.com/tristanmanchester/agent-skills/pull/7"},
{"directory":"textual-tui","status":"merged","present_in_main":true,"pr":31,"pr_url":"https://github.com/tristanmanchester/agent-skills/pull/31"},
{"directory":"todoist-api","status":"merged","present_in_main":true,"pr":11,"pr_url":"https://github.com/tristanmanchester/agent-skills/pull/11"},
{"directory":"track17","status":"merged","present_in_main":true,"pr":4,"pr_url":"https://github.com/tristanmanchester/agent-skills/pull/4"},
{"directory":"tracking-pettracer-location","status":"merged","present_in_main":true,"pr":33,"pr_url":"https://github.com/tristanmanchester/agent-skills/pull/33"},
{"directory":"wordly-wisdom","status":"merged","present_in_main":true,"pr":34,"pr_url":"https://github.com/tristanmanchester/agent-skills/pull/34"},
{"directory":"write-good-docs","status":"closed_unmerged","present_in_main":true,"pr":22,"pr_url":"https://github.com/tristanmanchester/agent-skills/pull/22"}
],
"merge_dependencies": [{"pr":10,"requires_pr":9,"status":"satisfied","pr9_merged_at":"2026-09-14T06:47:31Z","pr10_merged_at":"2026-09-14T06:48:46Z"}],
"latest_review": {
"record": "maintenance/pr-review-20260914.md",
"merged_skill_prs": [38,37,36,35,27,34,33,29,32],
"reviewed_heads": {
"38":"b62db4a709c54c607f9381e068f6c741944398b1",
"37":"e49742a3cd46425ca0789e4887401d9db1bb0c88",
"36":"5b3ce2ce891f6b452398c540a4295a79064c39a5",
"35":"c312a036cd29a84c39115a0da005ba38f57b0c0a",
"27":"6cb80972bf3ca25c84cbf96ccc5a93f874f5d2df",
"34":"fa3415399d1ee2cc91ab5e8c686a34fd12987404",
"33":"0b69400a081340e4de43173774cb0afe5ee604ab",
"29":"1d6cafb1695014e08d85b8d14a80261baa86023b",
"32":"fcf3cac2e0af023b11fa4d5e067ebdbf7db37dad"
},
"local_tests_passed": 36,
"suites": {"wordly_wisdom_python":12,"pettracer_python_mocked_transport":12,"skia_geometry_node":12},
"additional_checks": ["Direct --help execution of both decision entrypoints with Git executable modes","Strict TypeScript compilation of Skia geometry","Current review-thread, diff, status and check-run inspection"],
"ci": "No commit statuses or check-runs attached to these nine reviewed skill heads. Local tests are not CI.",
"limits": "No live provider/account calls, native-package type-check, device/browser/screen-reader tests, clinical/routing study, full retained-template execution, or complete rerun of earlier batches."
},
"historical_records": ["MAINTENANCE.md","maintenance/follow-up-20260914.md"],
"closed_pr_policy": "PR22 remains closed unmerged; it was not reopened or recreated. The original write-good-docs skill remains in main.",
"repository_housekeeping_pr": 25,
"status_authority": "GitHub PR state after this dated snapshot; publication of this file does not update installed skills."
}
maintenance/pr-review-20260914.md
+97
View File
@@ -0,0 +1,97 @@
# PR review and skill merges — September 14, 2026
## Outcome and scope
The recovery found ten open PRs. Nine were skill proposals; all nine have now
merged through normal squash merges guarded by the exact reviewed head SHA.
The interrupted run had already committed their skill-level bot fixes. This
continuation verified those changes, ran the applicable local regressions,
replied with evidence, and merged the ready proposals rather than recreating them.
At skill-merge completion, main is `2572ed26fee46da651431bb7a4aaf2b3c5a9ed5f`.
All 34 skill proposals have merged: 33 updates and the Meta control retirement.
PR22 remains closed unmerged. The two unchanged keep decisions remain intact,
leaving 36 active skill directories from the 37-directory baseline.
PR25 publishes the repository-only catalogue and records. Its refreshed README
corrects the Reddit name, retrieval model, and entry-point guidance; it removes
stale proposal-status wording. Its branch is reconciled against the completed
main tree while preserving the merged skill trees. The temporary
`pr-review-snapshot.yml` acquisition workflow is excluded from the shipped tree;
it is not a test suite or permanent CI.
## Reviewed skill heads and findings
| PR | Reviewed head | Finding checked and disposition |
| --- | --- | --- |
| [#38](https://github.com/tristanmanchester/agent-skills/pull/38) | `b62db4a709c5` | Linked form guidance now permits context-appropriate error summaries or first-invalid focus. Visual guidance removes universal shadow/border rankings and distinguishes elevation from focus. Merged. |
| [#37](https://github.com/tristanmanchester/agent-skills/pull/37) | `e49742a3cd46` | Workshop scores now distinguish assessed, unverified, and out-of-scope evidence instead of requiring numeric scores for unknowns. Merged. |
| [#36](https://github.com/tristanmanchester/agent-skills/pull/36) | `5b3ce2ce891f` | Boundary-lint example covers js/jsx/mjs/cjs/ts/tsx/mts/cts and calls for actual parser/resolver coverage checks. Merged after source review, not an ESLint execution. |
| [#35](https://github.com/tristanmanchester/agent-skills/pull/35) | `c312a036cd29` | Description explicitly includes people experiencing coercion or monitoring while excluding assistance to perpetrate it. Merged; no clinical or model-routing efficacy claim. |
| [#27](https://github.com/tristanmanchester/agent-skills/pull/27) | `6cb80972bf3c` | Note redirection and upload paths are quoted. Native CLI and memory guidance retain exact resource/account scope and no blind mutation replay. Merged. |
| [#34](https://github.com/tristanmanchester/agent-skills/pull/34) | `fa3415399d1e` | Both Python entrypoints have Git mode 100755 and launch directly. Numerical regressions passed. Merged. |
| [#33](https://github.com/tristanmanchester/agent-skills/pull/33) | `0b69400a0813` | Malformed position objects fail instead of becoming a successful no-fix response. History membership is inside/outside/unknown independently of freshness, with raw records preserved. Merged. |
| [#29](https://github.com/tristanmanchester/agent-skills/pull/29) | `1d6cafb16950` | Linked build troubleshooting no longer prescribes obsolete postinstall or broader lifecycle-script permissions. Changed gesture/snapshot examples were inspected and geometry tests passed. Merged. |
| [#32](https://github.com/tristanmanchester/agent-skills/pull/32) | `fcf3cac2e0af` | Read-only retrieval guidance reviewed. Its root-catalogue finding is repaired in companion repository PR25, preserving one skill per PR. Merged. |
The nine proposals had eleven inline bot findings in total. Ten skill-local
findings were verified and answered in their review threads before merging.
The remaining catalogue finding is addressed by the README in this repository
PR. Replies document the evidence; they do not imply that GitHub's formal
thread-resolution state or an independent re-review was changed.
## Tests actually executed in this continuation
| Suite | Passed | Scope |
| --- | ---: | --- |
| wordly-wisdom | 12 | Fixed-anchor utility, finite and exact inputs, scenario arithmetic, JSON bounds, failure status, and unchanged input models |
| tracking-pettracer-location | 12 | Position/time parsing, device selection, bounded mocked transport, malformed position types, and inclusive history-window boundaries |
| react-native-skia geometry | 12 | Strict-compiled pure geometry, fitting, successive zooms, focal invariance, clamping, invalid inputs, and a deterministic 500-update sequence |
Total: **36 distinct passing tests**. Both decision entrypoints also launched
with `--help` directly using the executable modes recorded in Git. Runtime was
Python 3.13.5, Node 22.16.0, and TypeScript 5.8.3 on Linux.
The tested files were reconstructed from the fetched source and patches rather
than a full clone. Git blob comparisons matched the three decision implementation
files, the PetTracer implementation and both test files, and the Skia geometry
module. Tests use synthetic data and explicit mocks where noted; no account
credentials or private user data were used.
There were **no commit statuses or check-runs attached to the nine reviewed skill
heads**. That is absence of CI, not a green CI result. Each head's current review
comments and changed content were inspected before the guarded merge.
No native-package type-check, Worklets transformation, device/browser rendering,
screen-reader test, real Textual/Pilot run, provider request, clinical study, or
controlled agent-output evaluation ran in this continuation. Earlier batches'
suites were not all rerun. These limits do not erase the local test results, but
they constrain what those results establish.
## Dependencies and preserved decisions
Meta CLI [#9](https://github.com/tristanmanchester/agent-skills/pull/9) merged at 06:47:31 UTC on September 14;
control retirement [#10](https://github.com/tristanmanchester/agent-skills/pull/10) merged at 06:48:46 UTC. The prerequisite
was present before deletion. PR22 was rechecked closed without merging and left
alone; its packaging proposal is not silently bundled into another PR.
The [September 13 source review](../MAINTENANCE.md) and
[early September 14 template follow-up](follow-up-20260914.md) remain historical
records of their own work and tests. Their earlier open-PR counts are superseded
for current status by this record and the
[checkpoint](modernisation-progress.json). GitHub is the authority for subsequent
changes. A merge updates repository source, not an existing local installation.
## Primary contracts rechecked
The specific changing facts used in review were checked against the relevant
first-party documentation, including:
- [Skia installation](https://shopify.github.io/react-native-skia/docs/getting-started/installation/): dependency-delivered binaries rather than a required postinstall downloader.
- [Reddit Data API access](https://support.reddithelp.com/hc/en-us/articles/16160319875092-Reddit-Data-API-Wiki) and [the August 5 platform announcement](https://redditinc.com/news/modernizing-reddits-infrastructure-and-moderation-tools): approved access and gradual transition, not an instantaneous universal shutdown.
- [W3C form notifications](https://www.w3.org/WAI/tutorials/forms/notifications/): the context-sensitive notification and focus patterns used in the linked form review.
- [ESLint configuration](https://eslint.org/docs/latest/use/configure/configuration-files): configuration scope is distinct from proving every import path is enforced.
The numerical and file-mode findings are supported by the inspected source and
local regressions. Documentation checks do not establish a particular application's
runtime behaviour or an account's access.