agentleFS
Zengxiansheng007Zengxiansheng007/skills-old/skills/task-understanding-router/SKILL.md

task-understanding-router

Skill1 starsChanged 3 months agoMITView on GitHub

Classify and visibly frame a new user task before execution, then route it to direct execution, one high-value clarification, an existing safety or approval gate, or grill-system. Use for new tasks and material goal changes; do not re-enter for ordinary follow-ups or active grill-system answers.

The file

We split this skill into the rules it gives an agent

10 rules. About 0 are common, 7 are distinctive, and 3 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈7

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 3

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Use this as an instruction-level router, not a function or a platform hook.

    line 8
  2. Distinctive:

    Before ordinary routing, check the full-grill trigger in routing contract.

    line 13
  3. Distinctive:

    Ask at most one clarification question.

    line 16
  4. Very specific:

    Do not recreate grill-system's question pack, recursively invoke grill-system, or show a router understanding block while full-grill-active.

    line 21
  5. Very specific:

    Treat a single-file, specified, reversible text edit with no runtime behavior change as eligible for brief.

    line 22
  6. Distinctive:

    Use the deterministic fixture validator only to validate assets.

    line 23
  7. Distinctive:

    Read routing contract for state, trigger, priority, and classification rules.

    line 27
  8. Distinctive:

    Read output contract before composing a visible response.

    line 28
  9. Distinctive:

    Read handoff contract before delegating to grill-system.

    line 29
  10. Very specific:

    Use assets/fixture-suite.json and scripts/validate_fixture_suite.py when changing rules or regressions.

    line 30

≈ “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

Task Understanding Router

Use this as an instruction-level router, not a function or a platform hook. It has no programmatic input/output API and cannot guarantee that every message is intercepted.

Workflow

  1. Determine the conversation state. Do not re-enter when the user is answering an active grill-system question or making an ordinary follow-up. Treat an explicit replacement such as “停止当前任务,改为 X” as a new task.
  2. Before ordinary routing, check the full-grill trigger in routing contract. If it matches, create the minimal handoff described in handoff contract, set the conceptual state to full-grill-active, and let grill-system own the question sequence.
  3. Otherwise select one mode, in priority order: safety-gated, detailed-clarify, detailed-execute, brief. Read routing contract for observable criteria.
  4. Produce only the visible structure required by output contract. Do not reveal hidden chain-of-thought. Mark assumptions as assumptions. For a valid silent ordinary-execution request, do not emit router/classification commentary in any user-visible channel.
  5. Ask at most one clarification question. After an answer, ask another only if the remaining unknown materially changes the deliverable, external action, irreversibility, or safety conclusion; otherwise state an assumption and proceed.

Boundaries

  • A user request to hide ordinary understanding does not bypass classification, clarification, safety, approvals, or an explicit full grill.
  • Do not recreate grill-system's question pack, recursively invoke grill-system, or show a router understanding block while full-grill-active.
  • Treat a single-file, specified, reversible text edit with no runtime behavior change as eligible for brief; escalate cross-file, behavior/config/API/data, unclear-scope, or hard-to-reverse work.
  • Use the deterministic fixture validator only to validate assets. It does not invoke a model or prove runtime discovery.

Resources

  • Read routing contract for state, trigger, priority, and classification rules.
  • Read output contract before composing a visible response.
  • Read handoff contract before delegating to grill-system.
  • Use assets/fixture-suite.json and scripts/validate_fixture_suite.py when changing rules or regressions. Record real-session results using assets/forward-test-record.schema.json.

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.