Scan Report
5 /100
proof-agent
Adversarial verification for AI-generated work using GitHub Copilot
A well-documented adversarial verification tool for AI-generated code with no malicious behavior detected.
Safe to install
This skill is safe to use. No action required.
| Resource | Declared | Inferred | Status | Evidence |
|---|---|---|---|---|
| Shell | WRITE | WRITE | ✓ Aligned | entrypoint.sh:1 — all shell execution is documented in SKILL.md |
| Network | READ | READ | ✓ Aligned | fact-check.sh:48,55,65 — HTTP checks to npm registry, GitHub, and target URLs; d… |
| Filesystem | READ | READ | ✓ Aligned | verify.sh:1 — reads git diffs only; entrypoint.sh:5 — reads verification_prompt.… |
| Environment | NONE | READ | ✓ Aligned | action.yml — reads standard GitHub Actions env vars (GITHUB_TOKEN, PR_NUMBER); n… |
| Skill Invoke | WRITE | WRITE | ✓ Aligned | SKILL.md:1 — explicitly spawns independent verifier subagent via gh copilot |
| Clipboard | NONE | NONE | — | No clipboard access found |
| Browser | NONE | NONE | — | No browser automation found |
| Database | NONE | NONE | — | No database access found |
1 findings
Medium External URL 外部 URL
https://x.com/acolombiadev CONTRIBUTING.md:146 File Tree
11 files · 37.4 KB · 1381 lines Markdown 3f · 572L
Shell 3f · 448L
Python 3f · 280L
YAML 1f · 66L
TOML 1f · 15L
├─
▾
proof_agent
│ ├─
__init__.py
Python
│ ├─
config.py
Python
│ └─
verifier.py
Python
├─
▾
scripts
│ ├─
fact-check.sh
Shell
│ └─
verify.sh
Shell
├─
action.yml
YAML
├─
CONTRIBUTING.md
Markdown
├─
entrypoint.sh
Shell
├─
pyproject.toml
TOML
├─
README.md
Markdown
└─
SKILL.md
Markdown
Dependencies 1 items
| Package | Version | Source | Known Vulns | Notes |
|---|---|---|---|---|
pyyaml | >=6.0 | pip | No | Version constrained, low-risk dependency |
Security Positives
✓ SKILL.md thoroughly documents all capabilities including network calls, shell execution, and subagent invocation
✓ All network requests target legitimate services: GitHub API, npm registry, and explicitly-requested URLs for fact-checking
✓ Python code is clean, readable, and uses standard library + pyyaml (pinned to >=6.0)
✓ No obfuscation, base64 encoding, or anti-analysis techniques found
✓ No credential harvesting or environment variable iteration for secrets
✓ No remote code execution from untrusted sources — curl calls are for verification only
✓ GitHub Actions token access is scoped to GitHub API operations (PR comments) as declared
✓ fact-check.sh is a pure verification utility with no side effects or data exfiltration
✓ Dependencies are minimal (pyyaml) and documented
✓ The skill follows a legitimate adversarial verification pattern (worker vs. verifier separation)