setup
Report (check, the default) or resolve (apply) the knowledge plugin's library_dir against the repository's artifact convention and its extraction prerequisites; apply install-deps provisions the extraction node deps and Chromium.
- License 许可
- MIT
- 最近更新
- 2026-11-15
- Stars
- ★ 30
- Forks
- 7
01 前言 Front Matter
- description
- Report (check, the default) or resolve (apply) the knowledge plugin's library_dir against the repository's artifact convention and its extraction prerequisites; apply install-deps provisions the extraction node deps and Chromium.
- argument-hint
- [check|apply] [install-deps]
- user-invocable
- true
- disable-model-invocation
- true
02 资源文件 1
- 参考文档evals/evals.json3.3 KB
03 SKILL.md 正文
Purpose
Bring the knowledge plugin to a working state: confirm the effective library_dir against this
repository's artifact convention, and verify (or provision) the extraction pipelines'
prerequisites. library_dir is a personal userConfig option that Claude Code prompts for and
stores; this skill reports where the effective value came from and never writes it.
- Pointer: when deciding where a
library_dirvalue is stored and which settings scopes are read for it, fetch https://code.claude.com/docs/en/plugins/manifest-reference#where-values-are-stored and https://code.claude.com/docs/en/settings-reference#pluginconfigs live; the option contract is https://code.claude.com/docs/en/plugins/manifest-reference#user-configuration.
Check-centric per the uniform setup contract (docs/plugin-philosophy.md
"Setup is explicit and repeatable" in the marketplace repository): check inspects and
reports, apply resolves what check found, and the extraction-dependency provisioning is a
distinct opt-in subaction. The library_dir option is Claude-Code-owned; this skill never
writes it, only reports and routes.
Action routing: no argument or check runs the check; apply runs the check first, then the
library_dir guidance; apply install-deps additionally provisions the extraction dependencies
below. All actions are non-interactive when the action is given, never prompt.
check (read-only)
Report a PASS/FAIL/INFO table with one remediation line per FAIL. Do not edit settings, write config, or install anything.
library_dirvs repository convention. Read the rendered${user_config.library_dir}value (default.= repository root). Inspect the consumer'sCLAUDE.md,AGENTS.md,.claude/rules, and existing artifact directories for a declared knowledge/artifact convention, the team source of truth. Do not inferlibrary_dirfrom the lifecyclememory_dir: the memory root holds lifecycle working documents, whilelibrary_dirowns the knowledge corpus. Mapping both to.workwould nest the video-digest pipeline's own.work/<watch-epic>/...layout as.work/.work/.... PASS when the personal value matches the convention (or the portable.default with no distinct convention). FAIL on a mismatch (remediation: Claude Code's plugin configuration prompt forknowledge) or a machine-absolute path (not portable for repository work).upstream_records_dir. INFO. Report the directory/knowledge:absorb-upstreamwrites to and its layer:upstream_records_dirin<home>/knowledge.yaml(<home>from the pointer region ofAGENTS.mdorCLAUDE.md, elsedocs/conventions), else the defaultdocs/upstream, under the root rule and value rules of${CLAUDE_PLUGIN_ROOT}/reference/config.md. An invalid value is a FAIL naming the file, key and value (remediation: edit the team file by hand;applydoes not write it).- Extraction node deps. INFO. Probe whether
${CLAUDE_PLUGIN_DATA}/node_modules/@melodic/video-digestionexists (the video-digest and course-digest pipelines share it). Missing is INFO, not FAIL: each ingest skill self-provisions on first run, andapply install-depspre-provisions. Remediation:apply install-deps. - Playwright Chromium. INFO. Probe whether the browsers path
(
PLAYWRIGHT_BROWSERS_PATH, else${CLAUDE_PLUGIN_DATA}/ms-playwright) holds achromium-*orchromium_headless_shell-*build (course-digest frame capture). Missing is INFO with remediationapply install-deps. - OS-level media tools. INFO. Probe
yt-dlp --version(video acquisition),ffmpeg -version(frame extraction, watch/course actions), andmagick -version(ImageMagick 7, contact sheets. Watch/course actions). Each absence is INFO with the platform install command from the ingest skill's Prerequisites; this skill never installs system packages.distill(PDF) needs none of these.
apply (idempotent)
Run check, then resolve each finding. Re-running after everything passes changes nothing and
reports "already configured".
-
library_dirmismatch. Guidance only.library_diris a personaluserConfigscalar; never hand-editpluginConfigsor write Claude Code settings. Reconfigurelibrary_dirthrough Claude Code's native flow, interactively with/plugin configure knowledge@<marketplace>or by a headless route, as the marketplace's plugin-reconfiguration convention gives them in § The routes (it owns the verified-version record). Apply the scope and failure-signal caveats in its § Caveats every setup skill's short form carries. Read both sections and print what they give, the fallback route, its behavior on an already-installed plugin and the scope to pass included, rather than only naming them, and leave out the verified-version record. Never uninstall to reconfigure; that section links what an uninstall deletes. Afterwards reruncheckin a fresh session, since a same-sessioncheckcan still show the old value, and report the observed effective value, never an unobserved change.- Pointer: for the headless command's flags, how it reports a saved or refused value, and when a saved value applies, fetch plugin CLI reference § plugin configure live.
For where
${user_config.*}resolves, see Plugin manifest reference: Reference a saved value. For a root outside the project and home directories, recommend the portable value forms from the README's option table (~prefix or${NAME}/%NAME%env-var reference) instead of a literal machine path, which guardrail hardcoded-path checks rightly block. -
apply install-deps. Provision the extraction dependencies. Only with this subaction, run the shared idempotent provisioner once per extraction package (each run installs that package's vendored node dependencies into${CLAUDE_PLUGIN_DATA}, and the course-digest run additionally provisions Chromium into${CLAUDE_PLUGIN_DATA}/ms-playwright):node "${CLAUDE_PLUGIN_ROOT}/vendor/node-launcher/setup-deps.mjs" --data-dir "${CLAUDE_PLUGIN_DATA}" --package "${CLAUDE_PLUGIN_ROOT}/skills/video-digest/extraction" node "${CLAUDE_PLUGIN_ROOT}/vendor/node-launcher/setup-deps.mjs" --data-dir "${CLAUDE_PLUGIN_DATA}" --package "${CLAUDE_PLUGIN_ROOT}/skills/course-digest/extraction"A stored fingerprint gates reinstalls, so re-running is safe and cheap. After provisioning, re-run the
checknode-deps and Chromium probes and report their actual results, never claim provisioned on the script exit codes alone. -
Invalid
upstream_records_dir. Guidance only. Name the file, key and value and the valid forms from${CLAUDE_PLUGIN_ROOT}/reference/config.md; never write the team file. -
OS-level media tools. Guidance only. For a missing
yt-dlp,ffmpeg, or ImageMagick 7, give the platform install command from the ingest skill's Prerequisites. This skill never installs system packages. -
Confirm. Report the observed
library_dir, the repository convention and any mismatch, and whether extraction dependencies were provisioned or intentionally skipped. Note that/knowledge:distillwrites to its explicitly named target skill rather than this setting.
Output
Report the observed personal value, the repository convention, any mismatch, the state of each
prerequisite, and the exact next action (/plugin configure knowledge for library_dir,
or apply install-deps for provisioning). A library_dir change takes effect in a fresh session,
not this one. Do not claim it this session.
Boundaries
- Do not run an ingestion pipeline. Provisioning dependencies is not the same as running one.
- Do not write the plugin cache, Claude Code user settings, or
pluginConfigs. - Do not invent organization-specific paths or environment-variable prefixes.
04 同仓库更多 Skills
track
Backlog reads through the bound tracker: search open and closed items, view one item with its comments and state history. Use when another skill searches or reads tracker items.
reviewer-media
Reviewer media for one pull request: decides whether it needs a narrated walkthrough, builds the storyboard from /review:explain-change's digest, has /explainer-video:produce render it and hands the manifest to /source-control:pull-request media. Use when asked for 'reviewer media', 'a walkthrough video for this PR', 'a produced video of this change', or when pull-request ready or monitor hands a PR over.
route
Model and effort per agent role (orchestrator, worker, verifier, retrieval) for a multi-agent run, resolved through the bundled defaults and the user, team and personal override config layers into fenced JSON with single-agent and fan-out variants and the layer behind each value. Use when asked 'which model for the subagents', 'route the roles' or 'model routing', or before a skill passes a role's model on a spawn.
route
Model and effort per agent role (orchestrator, worker, verifier, retrieval) for a multi-agent run, resolved through the bundled defaults and the user, team and personal override config layers into fenced JSON with single-agent and fan-out variants and the layer behind each value. Use when asked 'which model for the subagents', 'route the roles' or 'model routing', or before a skill passes a role's model on a spawn.
principles
Module-design vocabulary for architecture findings, and the checks built on it. Use when: 'what is a deep module', 'what is a seam', 'deep vs shallow module', 'what is the deletion test', 'leverage and locality', 'module design vocabulary', 'which word do we use for X in architecture findings'. Scanning a codebase for friction is /architecture:improve.
write
Structured five-field bug report (title, steps to reproduce, expected vs actual, justified severity, suggested fix location) from an informal description, read-only. Use when the user names a defect they observed, or asks for one written up ('report a bug', 'file a bug', 'write this up as a bug').