agentleFS
simscalsimscal/sapiens-dtj/.claude/skills/dispatch-agents/SKILL.md

dispatch-agents

Skill0 starsChanged 4 months agoMITView on GitHub

Use when implementing a user story, bug, or technical item across domains. Owns the parallel fan-out across backend/frontend/devops and collection of their structured results.

2 other projects use a version of this.

Adopted by

Most recently taken up by

  • Other versions2 projects use a version of this.

    Word for word · 0

    None found.

    Edited · 0

    None found.

    Wrote their own version · 2

    1. danielvm-git/bigpowers★ 292same name
    2. Nirusan/claude-config★ 7same name

    How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.

  • In this repository8 other files it gives its agents.

The file

We split this skill into the rules it gives an agent

9 rules. About 0 are common, 2 are distinctive, and 7 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 7

    Very specific

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

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

    Never read from external systems mid-run — everything an agent needs passed inline in its <context>.

    line 11
  2. Very specific:

    Every agent receives exactly one <decisions> block, sourced in strict order:

    line 40
  3. Very specific:

    When delta is provided, render <constraints> as:

    line 78
  4. Very specific:

    Never fetch, read, or re-derive context from inside spawned agents — pass everything inline via <context>.

    line 134
  5. Very specific:

    Never post comments, open issues, or notify any channel — return blocked reasons in $AGENT_RESULTS.

    line 135
  6. Distinctive:

    Never spawn agents sequentially — always one parallel message.

    line 136
  7. Very specific:

    Never cancel in-flight agents when one returns <blocked> — let them finish and collect.

    line 137
  8. Distinctive:

    Never emit empty optional XML sections — omit them.

    line 138
  9. Very specific:

    Never paraphrase ACs — pass the verbatim checklist from issue.body.

    line 139

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 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

Dispatch Agents

Flow: Pre-flight → Spawn → Collect. delta present → <constraints> carries it; agents implement only the delta rows (see Delta Rendering).

Never read from external systems mid-run — everything an agent needs passed inline in its <context>.

Inputs

ParameterRequiredNotes
issueyes{number, title, body} — the story or bug being implemented.
codebasesyes[{name, role, path}] — resolved from project-config.
agentsyesall / explicit domain list / from-codebases-field.
tdd_issueno{number, body} — TDD issue for the sprint. Absent for bugs without TDD.
design_contextnoFrontend design context — Storybook surface stories for the surfaces the story touches. Frontend-only.
decisionsno<decisions> payload (type + verbatim content) — investigation or plan. Replaces TDD when provided (see Decision Source Precedence).
deltano{satisfied, to_add, to_remove, to_rewrite, affected_files} — agents constrain work to the delta.

Pre-flight

Validate before spawning. Blocker → return immediately with status: precheck_failed + reason; don't spawn. Warning → continue, include in $AGENT_RESULTS.warnings.

CheckOn failure
issue.body contains ## Acceptance CriteriaBlocker — "issue missing acceptance criteria"
Resolved agent list non-emptyBlocker — "no agents resolved"
Every codebases entry has a role matching a resolvable agent (backend, frontend, devops)Warning — skip unmappable entries, list them
agents: from-codebases-field and the Affected Codebases bullet list is present/parseableBlocker — "affected codebases not parseable"
tdd_issue absent and decisions absent and issue not a bugWarning — "dispatching without TDD or decisions on a non-bug issue"
design_context provided but no frontend agent resolvedWarning — "design_context provided but no frontend agent dispatched"

Decision Source Precedence

Every agent receives exactly one <decisions> block, sourced in strict order:

  1. decisions present → use verbatim with the caller's type (investigation, plan, etc.).
  2. tdd_issue present → pass full tdd_issue.body verbatim, emit type="tdd".
  3. Neither → emit type="none". Acceptable for bugs with sufficient issue.body; surface a warning in $AGENT_RESULTS.

Context Protocol

Build one <context> XML block per agent. Same shape every domain; omit inapplicable sections (no empty <design_context/> or <constraints/> tags). <acceptance_criteria> = verbatim checklist from issue.body, never paraphrased.

<context>
  <requirements>
    <story>[work-item statement from issue.body]</story>
    <acceptance_criteria>
- [ ] [AC 1]
- [ ] [AC 2]
    </acceptance_criteria>
  </requirements>
  <decisions type="tdd|investigation|plan|none">
    [verbatim TDD slice, investigation fields, plan sections, or "none"]
  </decisions>
  <design_context>
    <surface slug="..." route="...">
      <story path="[relative path to Storybook story file]">
        [verbatim Storybook story file content]
      </story>
    </surface>
    <!-- repeat for each matched surface; omit entirely for backend, devops, and bug fixes -->
  </design_context>
  <constraints>
    [omit entirely if no delta; otherwise see Delta Rendering]
  </constraints>
</context>

Delta Rendering

When delta is provided, render <constraints> as:

<constraints>
  <delta>
    <satisfied>[file list]</satisfied>
    <to_add>[items]</to_add>
    <to_remove>[items]</to_remove>
    <to_rewrite>[items]</to_rewrite>
    <affected_files>[file list]</affected_files>
  </delta>
  <instruction>Implement the delta only — preserve satisfied work, add to_add items, revert to_remove items, rewrite to_rewrite items. Do not modify files outside affected_files unless a delta row explicitly requires it.</instruction>
</constraints>

Spawn

Spawn all resolved agents in one parallel message.

Collect

Each agent returns exactly one of:

ReturnMeaningSkill action
<result> with <files_changed>Domain implementation completeRecord status: complete
<no_work> with <reason>No work neededRecord status: no_work with the reason
<blocked> with <reason>Cannot proceedRecord status: blocked with the reason; do not post anywhere

One agent blocks → let others finish, collect and return all results. Every agent returns <no_work> → add warning "all agents reported no work — verify dispatch was appropriate".

Returns

$AGENT_RESULTS:

status: complete | partial | blocked | no_work | precheck_failed
results:
  - agent: backend|frontend|devops
    codebase: <name from codebases input>
    status: complete | no_work | blocked
    files_changed: [<paths>]   # present when status: complete
    reason: <string>           # present when status: no_work or blocked
warnings: [<string>, ...]      # pre-flight + collection warnings

Top-level status:

  • precheck_failed — pre-flight blocker; results empty.
  • complete — every agent returned <result>.
  • no_work — every agent returned <no_work>.
  • blocked — at least one <blocked>, no <result>.
  • partial — mixed: at least one <result> and at least one <blocked> or <no_work>.

Constraints

  • Never fetch, read, or re-derive context from inside spawned agents — pass everything inline via <context>.
  • Never post comments, open issues, or notify any channel — return blocked reasons in $AGENT_RESULTS.
  • Never spawn agents sequentially — always one parallel message.
  • Never cancel in-flight agents when one returns <blocked> — let them finish and collect.
  • Never emit empty optional XML sections — omit them.
  • Never paraphrase ACs — pass the verbatim checklist from issue.body.
  • Filesystem: read-only on issue, tdd_issue, project-config; no writes from this skill.

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.