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.

referenceadversarial-probing.md

≈2.2k tokens on demand. Your agent reads this file only when SKILL.md points to it.

Adversarial Probing Catalog

Purpose: Read this when you need probe selection, category coverage, risk assignment, or the ADVERSARIAL_PROBES payload.

Contents

Probe Contract

Adversarial probing searches for what the specification forgot, contradicted, or left unsafe. Run it after extracting criteria and before issuing any final verdict.

Probe ID Convention

PRB-{category_code}-{NNN}

Category codes:
- BND = Boundary
- OMS = Omission
- CTR = Contradiction
- IMP = Implicit
- NEG = Negative
- CNC = Concurrency

Each probe must include:

  • id
  • category
  • target_criterion or affected scope
  • description
  • spec_gap
  • risk
  • suggested_criterion
  • implementation_status when implementation evidence exists

Risk Classification

Risk Definition Effect on verdict
CRITICAL Likely data loss, security breach, or system failure Blocks CERTIFIED; requires remediation
HIGH Likely user-facing failure or broken workflow Blocks CERTIFIED; remediation plan required
MEDIUM Edge-case or expectation gap with clear impact Report and route for remediation
LOW Minor or rare behavioral gap Document for awareness

Probe Families

Boundary (BND)

Use for limits, extremes, and threshold behavior.

Probe focus Typical checks Example gap
Empty or null input null, empty string, missing field No required-field behavior specified
Numeric boundaries 0, negative, max, max+1, rounding No overflow or precision rule
Collection size empty list, single item, max size No behavior for zero results
Date and time leap day, timezone edge, midnight crossover No timezone definition
Encoding and files Unicode extremes, zero-byte file, oversized file No invalid-file handling

Omission (OMS)

Use for behavior the spec forgot to mention.

Probe focus Typical checks Example gap
Error messaging distinct failure paths No error text or error code rule
Loading and empty states loading spinner, zero results No empty-state behavior
Timeout and partial failure slow dependency, batch partial success No retry or recovery rule
Undo and notification user reversal, event notice No rollback or notification rule
Auditability logging, tracking, audit trail No recordkeeping requirement

Omission review prompts:

  1. What happens when the feature fails?
  2. What states can each entity enter?
  3. What happens before and after the user action?
  4. What happens if the dependency is unavailable?

Contradiction (CTR)

Use for conflicting requirements inside one spec or across related specs.

Probe focus Typical conflict
Cross-feature behavior One feature allows X, another forbids X
Priority and UX "Simple" conflicts with "cover all edge cases"
Permissions Global admin power conflicts with local restriction
Data model Create requires a field, update makes it optional
Timing model Real-time promise conflicts with batch-only process
Security vs convenience Seamless flow conflicts with strict re-authentication

Contradiction review steps:

  1. Group requirements by shared entity, workflow, or permission.
  2. Compare action, timing, and state assumptions.
  3. Flag any requirement that negates or weakens another.
  4. Keep the probe open until the contradiction is clarified or resolved.

Implicit (IMP)

Use for hidden assumptions that implementations often depend on.

Probe focus Typical assumption
Time and locale server timezone, locale-specific formatting
Auth state user is already logged in or verified
Network conditions always-online or low-latency behavior
Ordering and idempotency stable sort order, safe retry
Scale and retention data size, rate limits, retention period
Platform support browser, device, or OS assumptions

Assumption prompts:

  1. Who is assumed to use this?
  2. Where is it assumed to run?
  3. When does ordering or timing matter?
  4. How much volume or scale is assumed?
  5. What breaks if the assumption is false?

Negative (NEG)

Use for forbidden, invalid, or unauthorized behavior.

Probe focus Typical checks
Invalid input malformed JSON, injection payloads, missing fields
Unauthorized access other user's data, admin-only endpoint
Invalid transitions impossible workflow steps or expired resources
Duplicate or out-of-order actions double submit, retry before completion
Resource exhaustion too many requests, full disk, OOM
Integrity violations deleting parent with active children

Concurrency (CNC)

Use for race conditions, ordering, and parallel activity.

Probe focus Typical checks
Simultaneous updates two users edit the same resource
Read-after-write stale read immediately after mutation
Double submission rapid repeat action
Cache or queue ordering stale cache, reordered messages
Lost updates last-write-wins without merge rule
Session collisions same user on multiple devices

Execution Strategy

Priority Order

  1. CTR because contradictions can invalidate the entire spec.
  2. OMS because missing requirements become production defects.
  3. NEG because validation and authorization gaps are high-risk.
  4. BND because data limits often break integrity.
  5. IMP because hidden assumptions create latent failures.
  6. CNC because concurrency is narrower but expensive when missed.

Minimum Probes per Mode

Mode Minimum probes Coverage rule
FULL 12 Cover all 6 categories
ADVERSARIAL 24 Cover all 6 categories with deeper variants
AUDIT 6 Focus on OMS and CTR
EXTRACT 0 No probing

Review Sequence

  1. Map each probe to a criterion or spec section.
  2. Verify the claimed gap against the actual spec text.
  3. Assign risk based on business, security, and integrity impact.
  4. Add a measurable suggested_criterion.
  5. Feed open probes into the compliance report and remediation plan.

Output Format

Probe Template

PROBE:
  id: PRB-BND-001
  category: Boundary
  target_criterion: AC-XXX-001
  description: "Empty email input on login form"
  spec_says: "User submits email and password"
  spec_gap: "No specification for empty email handling"
  expected_behavior: "Validation error displayed"
  risk: MEDIUM
  suggested_criterion: "When email field is empty, display 'Email is required' error"

ADVERSARIAL_PROBES

ADVERSARIAL_PROBES:
  total: 18
  by_category:
    boundary: 4
    omission: 5
    contradiction: 1
    implicit: 3
    negative: 3
    concurrency: 2
  by_risk:
    critical: 1
    high: 4
    medium: 8
    low: 5
  probes:
    - id: PRB-OMS-003
      category: Omission
      target: AC-LOGIN-001
      description: "Login spec does not mention account lockout"
      spec_gap: "No lockout mechanism specified after failed attempts"
      risk: CRITICAL
      suggested_criterion: "After 5 failed login attempts, lock the account for 30 minutes"
      implementation_status: NOT_IMPLEMENTED

Quality Rules

  • Do not keep a probe if the spec already addresses the concern.
  • Do not invent a suggested criterion that is unrelated to the existing business intent.
  • Treat unresolved CTR probes as release-blocking until clarified.
  • Prefer one precise probe over several overlapping weak probes.
  • Keep probes measurable so they can become criteria or tests immediately.

Categories + Minimums (SKILL.md excerpt)

Probe ID convention: PRB-{category_code}-{NNN}

Category Code Focus
Boundary BND Limits, thresholds, extremes
Omission OMS Missing required behavior
Contradiction CTR Conflicting requirements
Implicit IMP Hidden assumptions
Negative NEG Forbidden or invalid paths
Concurrency CNC Parallel or ordering issues

Minimum Probes per Mode

Mode Minimum probes Coverage
FULL 12 All 6 categories
ADVERSARIAL 24 All 6 categories with deeper coverage
AUDIT 6 Focus on Omission + Contradiction
EXTRACT 0 No probing

Every probe output carries Probe ID, Category, Description, Spec Gap, Risk, Suggested Criterion.

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