All skills
simota avatar

/chisel

@965f4f9
by shingo imotasimota/agent-skills85 stars
15

Converting a supplied prompt into an executable specification: detects vague quality/quantity/explanation/style/design/technical/judgment wording, role and persona theater, and self-contradiction, then replaces each with a numeric bound, an observable behavior, or a scorable criterion — with a per-term ledger of what changed and what stayed open. Don't use for AI system design, RAG, or eval harnesses (Oracle), PRD/SRS authoring (Scribe), spec conformance verification (Attest), or SKILL.md normalization (Gauge).

Use this Skill: https://skilld.dev/gh/simota/agent-skills/chisel

This session only. Nothing lands on disk.

referenceautorun-schema.md

≈630 tokens on demand. Your agent reads this file only when SKILL.md points to it.

AUTORUN / Hub Schemas

Purpose: Chisel-specific _STEP_COMPLETE.Output and NEXUS_HANDOFF payload fields. The protocol itself (mode semantics, _AGENT_CONTEXT input, error handling) lives in _common/AUTORUN.md; the canonical handoff envelope lives in _common/HANDOFF.md. Read when: Emitting _STEP_COMPLETE at the end of an AUTORUN step, or responding to ## NEXUS_ROUTING.


_STEP_COMPLETE

_STEP_COMPLETE:
  Agent: Chisel
  Status: SUCCESS | PARTIAL | BLOCKED | FAILED
  Output:
    deliverable: "specified prompt + ambiguity ledger"
    artifact_type: "prompt_specification | ambiguity_ledger | role_decomposition | prompt_audit | specified_brief"
    parameters:
      detections: "[count]"
      dispositions: "quantify:N behavioralize:N criteria:N decompose:N audience:N condition:N date:N parameterize:N keep:N delete:N"
      unresolved_parameters: "[count]"
      length_delta: "[+/- lines vs source]"
  Validations:
    intent_preserved: "[yes | flagged]"
    exit_checklist: "[passed | failed:<item numbers>]"
    fabricated_numbers: "[none | flagged]"
  Next: [NextAgent] | DONE
  Reason: [Why this next step]

BLOCKED is the correct status when the source's own purpose admits two incompatible readings — every downstream translation would inherit the wrong one. Return it for the hub's GATE; never resolve it by guessing.


NEXUS_HANDOFF — Chisel-specific fields

Envelope per _common/HANDOFF.md. Surface these inline:

NEXUS_HANDOFF:
  Agent: Chisel
  Chisel_findings:
    recipe: spec | scan | role | audit | brief
    detections_by_class: { quality: N, quantity: N, explanation: N, style: N, design: N, technical: N, judgment: N, open: N }
    kept_open: [{ term: <text>, reason: executor-has-better-information | reversible | premature-formatting }]
    routed_to_enforcing_layer: [{ requirement: <text>, layer: validator | permission | retrieval | approval | human-review }]
    interpretations: [{ source: <text>, reading_chosen: <text> }]   # each becomes a hub DEC-n
    exit_checklist: pass | fail:<items>

interpretations is the load-bearing field for the hub: every reading chosen where the source admitted more than one is a judgment call made without the user, and belongs in the Decision Ledger.


brief Recipe Output

When invoked at the Nexus SPECIFY phase, return SPECIFIED_BRIEF instead of the four-section deliverable. Schema, gate, and injection rules → _common/nexus/reference/specify-phase.md.

Source: SKILL.md on GitHub

No alerts13d3 checks · Risk SAFE
  • Gen Agent Trust Hub13d

    The skill is a prompt engineering assistant designed to detect and resolve ambiguity in user instructions. It transforms vague language into explicit, executable specifications through a structured workflow. The analysis found no evidence of malicious patterns, data exfiltration, or unauthorized command execution. The skill correctly identifies its own logical boundaries by routing technical requirements like security protocols and spend caps to appropriate enforcing layers rather than relying on prompt-layer hardening alone.

  • Socket13d

    No alerts

  • Snyk13d

    Risk: LOW · No issues

Signed by skilld at 965f4f9. This ties the file your Agent reads to that commit on GitHub. It does not review the instructions.

Last checked against GitHub 3 days ago.

Activeupdated 2 weeks ago

README badge

README badge for simota/agent-skills/chisel