artifact-verification

Verifies specs, docs, manifests, and packaged builds that lack a canonical test gate.

2|Updated Jul 18, 2026
One-click install
npx skills add https://github.com/Arasz/ai-badger --skill artifact-verification-arasz
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: artifact-verification
Source: https://github.com/Arasz/ai-badger/tree/main/features/common/skills/artifact-verification
Command: npx skills add https://github.com/Arasz/ai-badger --skill artifact-verification-arasz

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? Non-code artifacts like specs, docs, manifests, and published packages often have no test suite, so agents assert "done" without evidence or re-run identical checks on identical bytes. This Skill provides disciplined verification workflows for those artifacts, including canonical checkers, ad-hoc verification scripts, and evidence tracking. ## Core Features & Use Cases - Canonical gate enforcement: Runs the workflow-defined checker (e.g. spec_holes.py for task specs) instead of inventing parallel checks, and interprets its exit codes and hole inventory correctly. - Install-protocol review: Reviews manual fresh-install verification protocols against a false-pass checklist covering silent degradation, sha-pinned asset fallbacks, stdout/stderr traps, cache provenance, and MCP stdio framing. - Merged-PR verification in installed builds: Verifies whether a packaged build contains a merged PR via binary probing and git tree comparison, avoiding the invalid squash-ancestry check. - Use Case: After merging a fix into a published dotnet global tool, verify the installed 1.0.9 build actually contains PR #55 by comparing git rev-parse <hash>^{tree} values rather than trusting version-hash ancestry. ## Quick Start Verify that the installed build of my packaged CLI tool actually contains the changes from the merged pull request, and report which check you used.

Frequently Asked Questions about artifact-verification

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I verify a spec or document that has no test suite?

Use the workflow's canonical checker first, such as spec_holes.py for task specs, and treat its exit code and hole inventory as the report. If the gate is mid-workflow red, run an ad-hoc script asserting content markers, structure counts, and the checker's expected output.

How do I check if an installed package contains a merged PR?

Probe the binary's --help or feature verbs first, then compare git trees with git rev-parse <hash>^{tree} versus the squash commit's tree. Squash-ancestry checks like git merge-base --is-ancestor are invalid because a squash merge creates a new commit.

Why does a fresh-install test pass even though the feature is broken?

The protocol likely exercises a degraded path: fresh state lacks the settings that enable the real feature, so the happy path silently falls back. Assert side-effect counters of the real path and confirm degradation warnings are absent from stderr.

Why does my string search in a compiled .NET DLL fail?

.NET assemblies store strings in two encodings: IL literals in the #US heap as UTF-16LE and attribute arguments in the #Blob heap as UTF-8. Search both encodings and confirm a positive control string of the same kind is findable.

Why does the verification reminder keep firing after I ran the checks?

The evidence hook records the terminal command's working directory as the root, so an internal cd inside your script does not re-root the event. Re-run the script with the terminal workdir set to the worktree containing the changed paths.

When should I not re-run a deterministic verification check?

When the file bytes are unchanged since the last verified state, one stat or shasum proving byte-identity plus a one-line statement is sufficient. Deterministic checks need at most one re-run; running them repeatedly on identical bytes adds no evidence.