wiki-curate

Pass

Audited by Gen Agent Trust Hub on Aug 24, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill is strictly read-only as described in the documentation and architecture sections. It analyzes local markdown files in ~/vault/ and generates reports in /tmp/ without performing network operations or modifying the analyzed corpus.
  • [COMMAND_EXECUTION]: The skill uses Bash to invoke a Python script (scripts/wiki-curate-scan.py). This is standard behavior for an orchestrator skill and operates within the local environment scope defined for the Research Vault.
  • [DATA_EXPOSURE]: While the skill reads sensitive-looking files (research concepts and project logs like CLAUDE.md), it only produces local reports and does not exfiltrate this data over the network.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes markdown files from the vault. While these files are technically untrusted data, the analysis is described as 'pure deterministic analysis' via a Python script rather than passing the content to an LLM for reasoning, which effectively mitigates indirect injection risks.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 24, 2026, 06:17 AM
Security Audit — agent-trust-hub — wiki-curate