メニュー
Audit database indexes TODO: expand this paragraph with the user phrases that should trigger the skill (e.g. "find unused code", "audit my logs"). The description is the primary matching signal Claude uses, so make it specific.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
Process the top Todo card on a GitHub Project (v2) board end-to-end: pull the board into a task-agent `tasks.yml`, run `task-agent` to open a PR for the top pending task, sync the PR back to the card, then wait for Copilot to review the PR and apply fixes for actionable comments. Use when the user says "work on the next task", "do the next board task", "pick up the next GitHub project item", "process the next card", "work on the next item in github project", "run the next board task", "drain the top todo", "process the top board todo", "auto-process the next card", or any phrasing that means "take the next pending item from a GitHub Project board and implement it". Composes `gh-project-sync` → `task-agent` → `gh-project-sync` → `copilot-review-fixer` (skill wrapper shipped by `task-agent`) via the `workflow` skill against the fixed YAML at `${CLAUDE_PLUGIN_ROOT}/workflows/auto-board-task.yaml`. All arguments are forwarded verbatim to the workflow runner; validation happens there and inside the composed sub-sk
Reads unresolved Copilot review comments on a pull request and applies code fixes for actionable ones. Thin skill wrapper around the `copilot-review-fixer` sub-agent — callable as a workflow step.
Reads a list of tasks in a yaml to be done in certain github repos. Agents clone the repo, do the task, commit the changes and create a PR. Not to be called automatically by Claude by any means.
Reconciler between a GitHub Project (v2) board and a task-agent unified `tasks.yml`. Use when the user asks to "sync the project board with tasks.yml", "reconcile the GitHub project", "pull Todo cards into the task queue", "push completed tasks back to the board", or supplies both a project URL/number and a tasks file. Arguments use `key=value` grammar: `tasks=<path>` (required), one of `project_url=<url>` OR `project_owner=<owner> project_number=<n>` (required), optional `default_repo=<owner/repo>` (for draft cards without a linked issue), and optional column name overrides `todo_column=…`, `in_review_column=…`, `done_column=…`, `wont_do_column=…` (default: `Todo` / `In Review` (falling back to `In Progress` if the board has no `In Review` column) / `Done` / `Won't Do`). Idempotent — running it any number of times converges to the same state.
Scan all dependencies for outdated versions and known security vulnerabilities. Uses npm audit / pip-audit / cargo audit / bundler-audit with grep-based fallbacks when dedicated tools are unavailable. Produces a prioritised, read-only report. Use when the user asks to "audit dependencies", "check for vulnerabilities", "find outdated packages", "run a dependency audit", "check npm audit", "scan for CVEs in dependencies", or anything about dependency security or freshness.
Run a multi-step skill workflow declared in a workflow YAML file. Use when the user asks to "run the workflow at <path>", "execute workflow <name>", "run the <name> workflow", "chain these skills", "compose <skill-a> and <skill-b>", or supplies a path to a `.yaml` workflow definition. Arguments follow a `key=value` grammar: the first positional value is the workflow file path, remaining `key=value` pairs are workflow inputs. The runner executes steps strictly in order, halts on the first non-zero exit, and writes a summary JSON to the workflow directory.
| name | index-audit |
| description | Audit database indexes TODO: expand this paragraph with the user phrases that should trigger the skill (e.g. "find unused code", "audit my logs"). The description is the primary matching signal Claude uses, so make it specific. |
| tools | Bash, Read, Grep, TaskCreate, TaskUpdate |
TODO: one-paragraph summary of what this skill does and the value it delivers. Be concrete — name the artefact it produces or the change it makes.
This skill is TODO: read-only / mutating. TODO: describe what it touches.
Before doing any work, call TaskCreate for each phase below. Call
TaskUpdate (status in_progress) when you begin a phase and TaskUpdate
(status completed) when you finish it.
TODO: replace with the first investigation step. Include the bash commands the agent should run — don't ask it to guess.
# TODO: example data-gathering command
TODO: replace with the implementation step.
TODO: replace with the verification / report step. Define the output format explicitly — show a concrete example below.
TODO: concrete example of what the report or file should look like.