przeprogramowani__10x-cli/CHANGELOG.md

Changelog

All notable changes to @przeprogramowani/10x-cli are documented in this file.

The format follows Keep a Changelog, and this project adheres to Semantic Versioning.

[Unreleased]

Changed

• 10x bench-kit is generally available. The TENX_CLI_EXPERIMENTAL opt-in is gone: init and update are registered unconditionally, appear in --help, and are documented in the README. The experimental_locked error envelope no longer exists (the gating module has been removed).

Added

• 10x sync — bulk download & update with change visibility. One command to download every unlocked lesson (--all) or refresh the ones you've already downloaded (default), with a report of what changed upstream. Each not-updated resource prints the exact 10x get … command to take it; --force takes all upstream updates over local edits, --dry-run previews without writing. Exit code is worst-outcome (1 if any lesson errored, else 0).
• Unchanged lessons are skipped without a download: the catalog now advertises a per-lesson contentHash (added in @przeprogramowani/10x-toolkit), compared digest-vs-digest against the value stored in the manifest at last apply. Older backends/manifests without a digest fall back to always-fetch.
• planBundle() — a pure, non-writing, non-prompting writer planner that classifies per-file actions + conflicts; applyBundle now consumes it so preview and apply can't diverge.
• Manifest lessons[].catalogContentHash (additive, optional) and LessonSummary.contentHash (optional) carry the per-lesson digest. Change visibility covers skills and prompts (configs are create-only; rules are sentinel-managed).

[1.0.0] - 2026-05-03

Breaking changes

This release upgrades to the directory-shaped skill bundle defined by the delivery API at @przeprogramowani/10x-toolkit. CLIs older than 1.0.0 cannot parse bundles produced after the API deploy, and 1.0.0 cannot consume the old bundle shape. Both repos must roll forward together.

• Skill bundles ship as full directories. A skill is now { name, files: [{ path, content, executable? }] } rather than { name, content }. Every file under the skill directory in @przeprogramowani/ai-artifacts (e.g. scripts/, references/) materializes on disk with its executable bit preserved when set. Running 10x get m1l1 now writes .claude/skills/<name>/SKILL.md, .claude/skills/<name>/scripts/check-context.sh (with +x), and any other files under the source skill directory.
• ArtifactResponse is a discriminated union. /api/artifacts/.../skills/:name returns { type: "skills", name, files, universalContent? }; prompts/rules/configs still return { type, name, content }. Code reading result.data.content for skills will not type-check.
• Manifest schema bumped to v2. files.skills is now a record keyed by skill name → { files: string[] } (relative paths under the skill dir). v1 manifests (skills as string[]) are read as null — treated as no-prior-state for one cleanup cycle, then the v2 manifest takes over on the next apply. No migration script.

Added

• Per-file removal within a retained skill. When upstream drops a file from an existing skill, 10x get removes the local copy and prunes the now-empty parent directory; the skill directory itself stays put for subsequent applies.
• Path-traversal guard for skill files. assertSafeSkillFilePath rejects empty paths, absolute paths, drive prefixes, .. segments, backslash separators, and Windows-reserved names (CON/PRN/...). The writer aborts before any filesystem mutation when a bundle is tampered.
• Print-mode notice for multi-file skills. 10x get ... --print --type skills [--name X] emits SKILL.md to stdout (so redirects like > /tmp/SKILL.md capture only that), and writes Note: skill "<name>" has N additional files not shown in --print: ...\nRun without --print to materialize all files. to stderr. JSON mode is unchanged; the full files[] array is in the response.
• Per-file actions in WriteResult.skills. result.skills[i] is now { name, files: [{ path, absolutePath, action }] } instead of { name, path, action }. The --json output reflects the same shape; consumers reading per-skill status need to walk the files array.

Changed

• ToolProfile.skillDir(name) is a new required field on every profile, returning the skill directory path (without the /SKILL.md suffix). The existing skillPath(name) is unchanged for callers that need the SKILL.md file specifically.
• tool-switch migration walks every file under each skill directory — migrateArtifacts and deleteArtifacts no longer assume a single file per skill.

Removed

• Nothing student-facing was removed. Old ArtifactWrite { name, path, action } entries in WriteResult.skills are replaced by the per-file SkillWrite shape described above.