agentsclimarketplace

Evidence boundary

Skill Gesondian/ai-collab-governance-skills/skills/evidence-boundary

Hermes Loop Engineering skills for self-improving AI delivery governance.

Install
npx -y skills add Gesondian/ai-collab-governance-skills --skill evidence-boundary

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Use when an AI delivery conclusion depends on partial, indirect, inherited, mock, local, visual, or otherwise limited evidence.

SKILL.md

2.5 KB, 513 tokens by cl100k_base, as published. Nobody here has run it

Evidence Boundary

Core Principle

Evidence is a boundary, not a slogan. A conclusion may only cover what the evidence directly proves.

Evidence Classes

EvidenceCan SupportCannot Support
Route reachableThe route loads under tested conditions.The workflow is usable or business-correct.
Single surface checkThat exact surface passed the stated assertion.All similar surfaces passed.
Mock data renderThe UI can render a synthetic shape.Real business rules, permissions, or data lifecycles work.
Static scanCertain text or code patterns are absent or present.Runtime behavior is correct.
Visual screenshotLayout at one viewport and data state.Keyboard flow, permissions, or hidden states.
Prior receiptWhat was previously asserted by that receipt.New or uncovered behavior.

Procedure

  1. Name the claim being made.
  2. List the evidence references.
  3. For each evidence reference, state:
    • scope covered,
    • assertion level,
    • fact source,
    • freshness,
    • blind spots.
  4. Downgrade any claim that exceeds the evidence.
  5. Record missing evidence before assigning acceptance or owner conclusions.

Supported Claim Levels

Claim LevelUse When
observedA concrete signal was seen, but no broader assertion is made.
route_reachableNavigation reached the tested route.
surface_checkedA named surface passed named assertions.
workflow_checkedA multi-step workflow passed with stated data and role assumptions.
accepted_scope_limitedAcceptance is bounded to listed dimensions only.
not_supportedEvidence is insufficient for the proposed claim.

Output Contract

claim:
evidence_refs:
covered_scope:
unsupported_scope:
missing_evidence:
safe_wording:
must_not_claim:

Common Mistakes

  • Treating reachability as usability.
  • Treating one successful example as class-wide proof.
  • Treating mock rendering as real data correctness.
  • Inheriting a conclusion from another receipt without checking its scope.
  • Saying a capability is shared merely because a shared-looking file exists.

Safe Wording Pattern

Use: "Evidence supports X under Y conditions; it does not yet support Z."

Avoid: "The feature works" when only one route, data shape, viewport, or assertion was checked.

What ships with it

Read from the repository

Just SKILL.md. No reference files, no scripts.

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.