agentleFS
synaptiaisynaptiai/agent-capability-standard/skills/perspective-validation/SKILL.md

perspective-validation

Skill6 starsChanged 13 days agoApache-2.0View on GitHub

Create a Perspective Validation Checklist (PVC) report for a change, workflow, schema, or policy. Use when performing socio-technical review, governance review, operational readiness review, or when a PR touches schemas/hooks/skills/tools and needs a PVC report.

The file

We split this skill into the rules it gives an agent

9 rules. About 0 are common, 4 are distinctive, and 5 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈4

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 5

    Very specific

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

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

    If the target is a PR or branch diff, identify changed files and whether they touch critical paths (schemas/, hooks/, skills/, …

    line 37
  2. Distinctive:

    Collect evidence: Read the minimum set of files needed to justify scores

    line 39
  3. Very specific:

    Prefer evidence anchors as file:path#Lline (or file:path:line if needed)

    line 40
  4. Distinctive:

    When reviewing a change set, include at least one evidence item for “what changed” and “what it affects”

    line 41
  5. Very specific:

    Write the report: Create a new file in docs/reviews/pvc/ named:

    line 49
  6. Very specific:

    Start from pvc_report_template.yaml, then fill it in completely

    line 51
  7. Distinctive:

    If it fails, fix the report until it passes

    line 54
  8. Very specific:

    Only write under docs/reviews/pvc/ unless the user explicitly asks for additional changes.

    line 148
  9. Distinctive:

    Do not include secrets/PII in evidence excerpts.

    line 149

≈ “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

Intent

Produce a PVC report (Perspective Validation Checklist) that scores the target against key non-type failure modes (human factors, governance, economics, adversarial security, loop stability, data governance, and legitimacy) with concrete evidence references and remediation actions.

Success criteria:

  • A valid PVC report is written to docs/reviews/pvc/ and passes python skills/perspective-validation/scripts/validate_pvc.py
  • Each non-N/A score is justified with at least one evidence reference
  • Any FAIL/PARTIAL includes a concrete remediation action (P0/P1/P2)

References:

  • CHECKLIST.md
  • pvc_report_template.yaml

Inputs

ParameterRequiredTypeDescription
targetYesstringWhat to review (file/dir path, workflow name, or PR/change description)
risk_tierNoenumlow | medium | high (defaults to medium if unclear)
domainNostringDomain context (e.g., general, manufacturing, healthcare)
deploymentNostringDeployment context (e.g., local-dev, customer-facing, safety-critical)

Procedure

  1. Scope the review: Confirm target, domain, deployment, and risk_tier
    • If the target is a PR or branch diff, identify changed files and whether they touch critical paths (schemas/, hooks/, skills/, grounded_agency/, tools/, spec/)
  2. Collect evidence: Read the minimum set of files needed to justify scores
    • Prefer evidence anchors as file:path#Lline (or file:path:line if needed)
    • When reviewing a change set, include at least one evidence item for “what changed” and “what it affects”
  3. Score the checklist: For each PVC test ID (HF-101 … ECO-901), assign one of:
    • PASS: explicitly satisfied with evidence
    • PARTIAL: acknowledged but missing enforcement/measurement/runbook detail
    • FAIL: absent
    • N/A: not applicable to this target (include a brief reason in assumptions or in an action note)
  4. Write the report: Create a new file in docs/reviews/pvc/ named:
    • YYYY-MM-DD_<short-slug>.pvc.yaml
    • Start from pvc_report_template.yaml, then fill it in completely
  5. Validate: Run python skills/perspective-validation/scripts/validate_pvc.py
    • If it fails, fix the report until it passes
  6. Summarize actions: Provide the top remediation actions (P0 first), with crisp “what to change” and “where it lives” guidance

Output Contract

Return a structured object:

pvc_report_path: string  # Path under docs/reviews/pvc/
target: string
context:
  domain: string
  deployment: string
  risk_tier: low | medium | high
scorecard:
  HF-101: PASS | PARTIAL | FAIL | N/A
  HF-102: PASS | PARTIAL | FAIL | N/A
  HF-103: PASS | PARTIAL | FAIL | N/A
  ORG-201: PASS | PARTIAL | FAIL | N/A
  GOV-202: PASS | PARTIAL | FAIL | N/A
  ECON-301: PASS | PARTIAL | FAIL | N/A
  ECON-302: PASS | PARTIAL | FAIL | N/A
  SEC-401: PASS | PARTIAL | FAIL | N/A
  SEC-402: PASS | PARTIAL | FAIL | N/A
  CTRL-501: PASS | PARTIAL | FAIL | N/A
  ASSUR-601: PASS | PARTIAL | FAIL | N/A
  ASSUR-602: PASS | PARTIAL | FAIL | N/A
  DG-701: PASS | PARTIAL | FAIL | N/A
  ETH-801: PASS | PARTIAL | FAIL | N/A
  ECO-901: PASS | PARTIAL | FAIL | N/A
actions:
  - id: string
    priority: P0 | P1 | P2
    fix: string
    artifact: doc | workflow | policy | hook | test | benchmark | code
    owner: string
confidence: number  # 0.0-1.0
evidence_anchors: array[string]
assumptions: array[string]

Example

Example: Review a workflow DSL change

Input:

target: "schemas/workflow_catalog.yaml (new gates + recovery semantics)"
risk_tier: medium
domain: general
deployment: customer-facing

Output:

pvc_report_path: "docs/reviews/pvc/2026-01-29_workflow-catalog-gates.pvc.yaml"
target: "schemas/workflow_catalog.yaml (new gates + recovery semantics)"
context:
  domain: "general"
  deployment: "customer-facing"
  risk_tier: "medium"
scorecard:
  HF-101: PARTIAL
  HF-102: FAIL
  HF-103: PARTIAL
  ORG-201: FAIL
  GOV-202: FAIL
  ECON-301: FAIL
  ECON-302: N/A
  SEC-401: PARTIAL
  SEC-402: PARTIAL
  CTRL-501: PASS
  ASSUR-601: PARTIAL
  ASSUR-602: FAIL
  DG-701: FAIL
  ETH-801: N/A
  ECO-901: PASS
actions:
  - id: "DOC-001"
    priority: P0
    fix: "Add an operator-facing incident view schema and a minimal tabletop exercise for this workflow family."
    artifact: doc
    owner: "maintainers"
confidence: 0.7
evidence_anchors:
  - "file:schemas/workflow_catalog.yaml#L1"
assumptions:
  - "No runtime UI exists; report assumes log-based incident response."

Safety Constraints

  • This skill is intended to be read-mostly.
  • Only write under docs/reviews/pvc/ unless the user explicitly asks for additional changes.
  • Do not include secrets/PII in evidence excerpts; prefer references/hashes over raw content.

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.