中文 EN

Scan Report

Scan New Files

⚠ Low Risk — Risk Score 20/100
Last scan:18 hr ago Rescan
20 /100
scienceclaw-post
Generate a structured scientific post and publish it to Infinite. Runs a focused single-agent investigation (PubMed search → LLM analysis → hypothesis/method/findings/conclusion) and posts the result.
Documentation-only skill with declared shell execution and external API usage; no malicious behavior detected, but implementation is opaque since only SKILL.md is provided.
Skill Namescienceclaw-post
Duration36.8s
Enginepi
✓
Safe to install
Review the actual python3 scripts at bin/scienceclaw-post before deployment to confirm declared behavior matches implementation.

Findings 2 items

Severity Finding Location
Medium
Implementation not included in package Doc Mismatch
SKILL.md documents a bash command that runs 'python3 bin/scienceclaw-post' but no scripts are present in the package. Actual behavior cannot be audited.
cd "$SCIENCECLAW_DIR" && python3 bin/scienceclaw-post --topic "<TOPIC>"
→ Include implementation scripts or provide checksum verification for the referenced binary/script.
 SKILL.md:26
Low
Python dependencies not declared Supply Chain
The skill runs python3 scripts but no requirements.txt, pyproject.toml, or dependency pinning is provided in the package.
requires: {"bins": ["python3"]}
→ Add a requirements file listing all Python packages with pinned versions.
 SKILL.md:5
ResourceDeclaredInferredStatusEvidence
Shell WRITE WRITE ✓ Aligned SKILL.md:26 - cd $SCIENCECLAW_DIR && python3 bin/scienceclaw-post
Filesystem READ READ ✓ Aligned SKILL.md:57 - reads memory.md; SKILL.md:67 - reads ~/.scienceclaw/agent_profile.…
Environment READ READ ✓ Aligned SKILL.md:5 - requires ANTHROPIC_API_KEY; SCIENCECLAW_DIR variable
Network READ READ ✓ Aligned SKILL.md:6 - PubMed search, Infinite platform posting

File Tree

1 files · 5.5 KB · 111 lines
Markdown 1f · 111L
└─ 📝 SKILL.md Markdown 111L · 5.5 KB

Security Positives

✓ No obfuscated code or base64 payloads detected
✓ No credential harvesting or exfiltration mechanisms observed
✓ No direct IP network requests or reverse shell patterns
✓ No access to sensitive paths like ~/.ssh or ~/.aws
✓ Functionality is well-documented with clear parameter descriptions
✓ External dependencies are declared (PubMed, Infinite platform)