All skills
simota avatar

/attest

@c805268
by shingo imotasimota/agent-skills85 stars
15

Verifying spec compliance: extracts ACs from specs, adversarially checks conformance, generates BDD scenarios and traceability matrices. Use when impl must be proven to match a PRD/SRS/AC.

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

This session only. Nothing lands on disk.

referenceverification-methods.md

≈1.5k tokens on demand. Your agent reads this file only when SKILL.md points to it.

Verification Methods

Purpose: Read this when choosing static verification techniques, assigning per-criterion verdicts, formatting evidence, or deciding when to route to runtime verification.

Contents

Verification Approach

Attest performs static verification. It does not execute the system.

What Attest Can Verify

Aspect Method Confidence tendency
Feature existence Code search HIGH
Data flow correctness Logic trace MEDIUM
Error handling presence Error-path review HIGH
State transitions State check MEDIUM
API contract shape Signature matching HIGH
Business-rule implementation Logic trace MEDIUM
Configuration compliance Config inspection HIGH

What Requires Runtime Verification

Aspect Why static verification fails Route to
Performance thresholds Requires execution timing Bolt / Siege
Visual rendering Requires rendered UI Voyager / Echo
External integrations Requires live endpoints Voyager

Mark these as NOT_TESTED with a runtime verification plan.

Verification Methods

Method Purpose Typical evidence
CODE_SEARCH Confirm an implementation artifact exists Route, handler, component, or config reference
LOGIC_TRACE Verify business-rule ordering and transformations Call chain or condition chain
STATE_CHECK Compare state machine behavior to the spec Enum, reducer, guard, workflow mapping
ERROR_PATH Verify specified failure modes Error handler, validation, response path
ABSENCE_CHECK Confirm no implementation exists Search locations checked and no relevant hit

Must-Keep Absence Pattern

Criterion: AC-LOGIN-007 "Account lockout after 5 failed attempts"
Search locations:
  - src/controllers/auth.ts
  - src/middleware/auth.ts
  - src/models/user.ts
Result:
  No lockout logic found
Verdict:
  FAIL

Verdict Assignment

Per-Criterion Verdicts

Verdict Meaning Evidence requirement
PASS Implementation fully satisfies the criterion At least one direct code reference with matching logic
PARTIAL Implementation exists but misses required aspects Code reference plus explicit gap
FAIL Implementation contradicts or omits the criterion Contradiction evidence or confirmed absence
NOT_TESTED Static verification is insufficient Reason plus runtime test recommendation
AMBIGUOUS The spec is too vague to judge Reference to AMBIGUOUS_FLAG

Decision Flow

1. Can the criterion be verified statically?
   -> No: NOT_TESTED
   -> Yes:
      2. Does implementation evidence exist?
         -> No: FAIL
         -> Yes:
            3. Does it match the specification fully?
               -> Fully: PASS
               -> Partially: PARTIAL
               -> Contradicts: FAIL

Evidence Format

EVIDENCE:
  criterion_id: AC-LOGIN-001
  verdict: PASS
  method: CODE_SEARCH + LOGIC_TRACE
  references:
    - file: src/routes/auth.ts
      line: 42
      snippet: "router.post('/api/auth/login', authController.login)"
      relevance: "Endpoint definition matching spec"
    - file: src/controllers/auth.ts
      line: 78-95
      snippet: "async login(req, res) { ... }"
      relevance: "Handler validates email + password"
  gaps: []
  notes: "Implementation matches spec."

Gap Description for PARTIAL / FAIL

EVIDENCE:
  criterion_id: AC-LOGIN-007
  verdict: FAIL
  method: ABSENCE_CHECK
  references: []
  gaps:
    - type: MISSING_IMPLEMENTATION
      description: "Account lockout mechanism not found"
      spec_requirement: "Lock account after 5 failed login attempts"
      searched_locations:
        - src/controllers/auth.ts
        - src/middleware/auth.ts
        - src/models/user.ts
      impact: CRITICAL
  notes: "No login attempt tracking or lockout logic exists in codebase"

Confidence Scoring

Confidence Range Meaning
HIGH 0.8-1.0 Direct and unambiguous evidence
MEDIUM 0.5-0.8 Likely correct but requires some reasoning
LOW 0.2-0.5 Indirect evidence or heavy inference

Rule: when confidence < 0.5, mark the result as NOT_TESTED with a runtime recommendation.

Resource Allocation Guideline

Priority Effort share Minimum scope
CRITICAL 40% All five methods plus adversarial probing
HIGH 30% Inspection, logic analysis, and basic probes
MEDIUM 20% Inspection and lighter probing
LOW 10% Inspection only

When scope must be reduced, finish all CRITICAL criteria first, then HIGH, then sampled MEDIUM. Defer LOW to AUDIT mode before reducing higher priorities.

Static Methods + Runtime-Only Areas (SKILL.md excerpt)

Method Purpose
CODE_SEARCH Confirm implementation artifacts exist
LOGIC_TRACE Follow data and business-rule flow
STATE_CHECK Verify state transitions match the spec
ERROR_PATH Verify specified failure behavior
ABSENCE_CHECK Confirm a criterion has no implementation evidence

Runtime-Only Areas

Route these to NOT_TESTED with a runtime plan:

  • Performance thresholds
  • Concurrency behavior
  • Visual rendering
  • External API integration
  • UX quality

Source: SKILL.md on GitHub

1 alert13d4 checks · Risk CRITICAL
  • Gen Agent Trust Hub13d

    The 'attest' skill is a specification compliance tool that uses static analysis to verify code against requirements. While automated scanners flagged a blacklisted domain and one documentation file, these are confirmed to be benign placeholders and standard instructional content. The skill possesses an indirect prompt injection surface inherent to its function, but this is mitigated by strict evidence-based verification rules and adversarial probing requirements. No malicious behaviors, obfuscation, or unauthorized operations were detected.

  • Socket13d

    No alerts

  • Snyk13d

    Risk: LOW · No issues

  • ZeroLeaks5mo

    Score: 93/100 · 2 sections analyzed

Signed by skilld at c805268. This ties the file your Agent reads to that commit on GitHub. It does not review the instructions.

Last checked against GitHub 2 days ago.

Activeupdated 2 weeks ago

README badge

README badge for simota/agent-skills/attest