agentleFS
chipweaverchipweaver/veripower/skills/specification/SKILL.md

specification

Skill10 starsChanged 53 days agoMITView on GitHub

Establish requirements, design decisions, interfaces, clocks and constraints from the delivered intent, or revise them after evidence-based clarification; not for implementing RTL or verification.

5 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

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

  • 0

    Common

    Most of the 5 edited versions we read keep these.

  • 0

    Distinctive

    This author's own choices. Worth a look.

  • 2

    Very specific

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

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

    Read the evidence needed to resolve a question.

    line 10
  2. Very specific:

    Use check-hints-task-contract.md to describe the observations used by simulation.

    line 45

“Common” is measured: kept by at least 6 in 10 of the 5 edited versions of this file we could read. “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

Specification

Establish the task's requirements and implementation boundary from the original intent. Read {workdir}/dispatch.json; intent names the delivered material and its referenced authorities. Read the evidence needed to resolve a question. Write this stage's artifacts under {workdir}. Author directly or delegate focused work, retaining responsibility for the resulting decisions.

Establish or revise the definition

Continue from existing artifacts. Use changed inputs, failure records and actual user decisions to scope work; absence of caused_by does not imply a first delivery. Preserve unaffected content and stable requirement IDs. Read an upstream request's evidence before accepting its attribution.

Produce the requirements ledger using ledger-task-contract.md and the boundary using decompose-task-contract.md. design.md records design choices and joint obligations; the RTL author decides the module split. The boundary sidecars are manifest.json, clocks.json and top-io.json; their formats are in the corresponding reference schemas.

Check the ledger:

python3 <skill>/scripts/spec/__main__.py check-ledger --workdir {workdir}

<skill> is this skill's directory. The output shows unresolved/unassignable, outside-scope and rows requiring decisions, plus numerical targets. Resolve these according to the task and actual authorization. Explain choices requiring a human decision before seeking it; act within an existing delegation when permitted. Record material decisions and their basis in spec-review/decisions.md, identifying who decided and under what authorization. An agent decision is not a user quotation.

Distinguish a mistaken transcription or design assumption from a proposed change to acceptance. Correct the former from source evidence. A genuine acceptance change needs authorization and a clear account of what is no longer being claimed. Neither an approval nor a decision record proves reachability, physical feasibility or successful verification. Investigate factual questions or return them to the responsible stage rather than redefining a requirement to make a result pass.

Complete and review

Use check-hints-task-contract.md to describe the observations used by simulation. Check the ledger/boundary/hint references and derive constraints:

python3 <skill>/scripts/spec/__main__.py check-crossrefs --workdir {workdir}
python3 <skill>/scripts/spec/__main__.py derive-constraints --workdir {workdir}

These checks identify inconsistent references and invalid tool inputs; they do not establish the truth of an engineering assumption. Repair the side that is wrong based on evidence.

Have the complete delivery independently reviewed using review-task-contract.md. Assess findings against their sources, repair defects, and recheck affected artifacts. Have substantive changes reassessed. Retain the review and material decisions; assess acceptance impact independently of a finding's label. Present design.md, review paths and any unresolved decision to the user as needed.

Close

Finalize when this stage's work is complete or an unresolved issue must return to the caller:

python3 <skill>/scripts/spec/__main__.py finalize --workdir {workdir} \
  [--fail-reason "<unresolved cause and evidence>"]

Finalize checks current references, derives constraints, validates the ledger and rejects unresolved unassignable rows. It does not adjudicate review prose. Supply --fail-reason for an unresolved blocking issue or incomplete review; stage_specific.fail_reason carries that cause.

Exit 0 means result.json was written, pass or fail; return STATUS: DONE and let the caller route it. A nonzero exit means closure failed; resolve the cause or return STATUS: BLOCKED <cause>.

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.