agentleFS
websublimewebsublime/mister-anderson/skills/do/SKILL.md

do

Skill3 starsChanged 7 months agoMITView on GitHub

Dispatch the implementation supervisor for a bead task. Resolves the correct supervisor from the assignee field, checks branch state, and dispatches with full context.

The file

Show its rules and text

We split this skill into the rules it gives an agent

19 rules. About 0 are common, 10 are distinctive, and 9 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈10

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 9

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    If the user provides $ARGUMENTS, check if it contains a bead ID (e.g., bd-a3f, bd-001.2).

    line 14
  2. Distinctive:

    If bead ID provided: use it directly.

    line 16
  3. Very specific:

    Run bd ready --json to get unblocked tasks

    line 19
  4. Very specific:

    Investigation should exist: Read bead comments: bd comments {BEAD_ID}.

    line 26
  5. Distinctive:

    No investigation found for this bead.

    line 34
  6. Very specific:

    Ask: "Do you want to run /investigate {BEAD_ID} first, or proceed without investigation?"

    line 37
  7. Very specific:

    If user wants investigation → stop and recommend /investigate

    line 38
  8. Distinctive:

    If user proceeds → continue to dispatch </on-check-fail>

    line 39
  9. Distinctive:

    Read the design doc content — this is the implementation contract

    line 49
  10. Very specific:

    Read the assignee field from the bead JSON — this contains the supervisor name (e.g., rust-supervisor)

    line 54
  11. Very specific:

    Verify the agent file exists: .claude/agents/{assignee}.md

    line 56
  12. Distinctive:

    If found → supervisor resolved

    line 57
  13. Distinctive:

    If NOT found → warn user, fall through to manual selection

    line 58
  14. Distinctive:

    Ask: "Which supervisor should handle this task?"

    line 63
  15. Very specific:

    If user wants to create → stop and suggest /add-supervisor, then re-run /do

    line 64
  16. Distinctive:

    Ask: "Continue on the existing branch, or create a fresh branch?"

    line 74
  17. Distinctive:

    Ask: "Which branch should I base from?

    line 78
  18. Distinctive:

    Decisions/Deviations: if the supervisor logged DECISION or DEVIATION comments, explain each in plain language — what choice was made, why, …

    line 107
  19. Very specific:

    After the supervisor completes, recommend /review {BEAD_ID} for code review.

    line 113

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Do

Dispatch the implementation supervisor for a bead task. Handles supervisor resolution, branch management, and dispatch with full investigation context.


<on-init> If the user provides $ARGUMENTS, check if it contains a bead ID (e.g., `bd-a3f`, `bd-001.2`).

If bead ID provided: use it directly.

If no bead ID provided:

  1. Run bd ready --json to get unblocked tasks
  2. Present the list showing: ID, title, priority, and labels
  3. Pick by order and priority; if unclear, ask the user</on-init>
<on-check> 1. **Bead must exist.** Validate with `bd show {BEAD_ID} --json`. 2. **Investigation should exist.** Read bead comments: `bd comments {BEAD_ID}`. Search for `INVESTIGATION:`. </on-check><on-check-fail if="bead"> Bead does not exist. Inform the user and stop. </on-check-fail><on-check-fail if="investigation"> No investigation found for this bead. Warn: "No investigation found. The supervisor will implement without codebase analysis. This can lead to missed edge cases."

Ask: "Do you want to run /investigate {BEAD_ID} first, or proceed without investigation?"

  • If user wants investigation → stop and recommend /investigate
  • If user proceeds → continue to dispatch</on-check-fail>
<on-execute>

Step 1: Read Full Bead Context

  1. Parse bead JSON. Extract: description, acceptance, design, notes, status, parent, labels
  2. If bead is an epic child (BEAD_ID contains a dot):
    • Extract EPIC_ID, read the epic, extract design doc path
    • Read the design doc content — this is the implementation contract
  3. Store all context for the dispatch prompt

Step 2: Resolve Supervisor

  1. Read the assignee field from the bead JSON — this contains the supervisor name (e.g., rust-supervisor)
  2. If assignee is set and non-empty:
    • Verify the agent file exists: .claude/agents/{assignee}.md
    • If found → supervisor resolved
    • If NOT found → warn user, fall through to manual selection
  3. If assignee is empty or unset:
    • List available supervisors: find all *-supervisor.md in .claude/agents/
    • Present the list
    • Inform: "If the supervisor you need is not listed, create one with /add-supervisor {technology}."
    • Ask: "Which supervisor should handle this task?"
    • If user wants to create → stop and suggest /add-supervisor, then re-run /do

Step 3: Check Branch State

git branch -a | grep {BEAD_ID}

If branch exists:

  • Inform user: "A branch already exists: {branch-name}. This may be from a previous implementation attempt."
  • Ask: "Continue on the existing branch, or create a fresh branch?"
  • Include context in dispatch prompt

If no branch exists:

  • Ask: "Which branch should I base from? (default: main)"
  • Store as {BASE_BRANCH}

Step 4: Dispatch Implementation Supervisor

  1. Announce before dispatching:
    Dispatching {resolved-supervisor} for {BEAD_ID}: "{bead title}"
  2. Dispatch using exactly these parameters — no more, no less:
    Agent(
        subagent_type="{resolved-supervisor}",
        prompt="Implement BEAD {BEAD_ID}. [Include EPIC_ID if epic child]. Base branch: {BASE_BRANCH} — run `git checkout {BASE_BRANCH}` before creating your feature branch. Read the bead (bd show {BEAD_ID}) and comments (bd comments {BEAD_ID}) for full context — description, acceptance criteria, design notes, and investigation findings."
    )
    Do NOT add extra parameters unless the user explicitly requests it.
  3. The PreToolUse hook automatically injects the discipline reminder because the agent name ends in -supervisor.
</on-execute><on-complete> Present the result following the **Presenting to the User** guidelines:
  1. Headline: "{bead title} — implementation complete" in plain language
  2. What was done: 1-2 sentences summarizing what the supervisor built — translate the COMPLETED bead comment into natural language, don't paste raw
  3. Branch and PR: show branch name and PR URL (look for PR: in the completion report or bead comments)
  4. Decisions/Deviations: if the supervisor logged DECISION or DEVIATION comments, explain each in plain language — what choice was made, why, and what the alternative was
  5. Next step: recommend /review {BEAD_ID}
  6. Technical details: raw COMPLETED comment and file list in a <details> block</on-complete>
<on-next> After the supervisor completes, recommend `/review {BEAD_ID}` for code review. </on-next>

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.