agentleFS
alphazedealphazede/bearing-lite/skills/light-implementer/SKILL.md

light-implementer

Skill3 starsChanged 3 months agoApache-2.0View on GitHub

Execute one approved light slice exactly as its packet states, verified by the packet's deterministic command. Use for Light Implementer, light, mechanical, scaffold, bind, assemble, regenerate, or runbook packets. Do not use for judgement work, authoring, repair, review, or any decision.

The file

We split this skill into the rules it gives an agent

7 rules. About 0 are common, 6 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈6

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Use only when enabled in the selected profile and assigned the mechanical slice.

    line 12
  2. Distinctive:

    Match: every criterion holds:

    line 23
  3. Distinctive:

    No decision: no value chosen, ambiguity resolved, candidate selected, requirement, contract, rationale, or verification case written.

    line 29
  4. Distinctive:

    Do exactly what the packet states, inside the write set, and nothing else.

    line 40
  5. Distinctive:

    Run the packet's verification command at the candidate revision.

    line 41
  6. Very specific:

    If any step needs a choice the packet did not make, stop before writing further and return NEEDS_MORE_EVIDENCE with reclassify: judgement …

    line 44
  7. Distinctive:

    Never author, repair, self-certify, expand the write set, or publish.

    line 56

≈ “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

Light Implementer

Use only when enabled in the selected profile and assigned the mechanical slice. Inputs fully determine the output, a command decides pass or fail, and nothing is decided in-session. Loop: execute exact packet → deterministic self-check → return evidence or typed failure.

Inputs and match

  • Inputs: the Implementer packet contract (baseline, objective, exact write set, authority, commands, stop rule, return schema, visible wave receipt, profile identity from the recorded snapshot) for a slice whose work_class is light.
  • Match: every criterion holds:
    1. Inputs are all named and present: paths, digests, UIDs, a runbook. Nothing is discovered or interpreted.
    2. The transformation is mechanical: copy, assemble, format, fill a template, append rows, run a documented command, record its output.
    3. The packet names the command whose exit status verifies the output.
    4. No decision: no value chosen, ambiguity resolved, candidate selected, requirement, contract, rationale, or verification case written.
    5. Bounded blast radius: one write set, no product source, no host or runtime mutation beyond a documented read-only or prepare call.
  • Non-match: a judgement slice; a packet whose LOOP says repair your own findings; an [OPEN] value; a register conflict to weigh.

Algorithm

  1. Revalidate the checkout lease exactly as the Implementer does. On mismatch return WAITING_ON without writing.
  2. Do exactly what the packet states, inside the write set, and nothing else.
  3. Run the packet's verification command at the candidate revision. Record exit status, output digest, and changed paths. Any deterministic-verification receipt is diagnostic only and never satisfies an assurance gate.
  4. If any step needs a choice the packet did not make, stop before writing further and return NEEDS_MORE_EVIDENCE with reclassify: judgement and the exact question. Never guess, never escalate silently.
  5. There is no in-wave repair loop. A failing command returns the typed failure with its output; the Orchestrator decides.

Return and recovery

Return CANDIDATE_READY, NEEDS_MORE_EVIDENCE, WAITING_ON, or OWNER_DECISION_REQUIRED with verdict, candidate_ref, changed_paths, tests, findings, blocker, and reclassify when set.

Never author, repair, self-certify, expand the write set, or publish.

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.