wiki-retrieve

Pass

Audited by Gen Agent Trust Hub on Jul 30, 2026

Risk Level: SAFEDATA_EXFILTRATIONPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [DATA_EXFILTRATION]: The skill provides an optional feature to send local document content to the Anthropic API to generate contextual prefixes. This egress is documented and gated by the --allow-egress flag and setup-time user prompts. Anthropic is a trusted organization.
  • [PROMPT_INJECTION]: The skill ingests local markdown files from the wiki/ directory to create a retrieval index. This introduces a surface for indirect prompt injection if the source documents contain malicious instructions.
  • Ingestion points: Markdown files in the wiki/ directory.
  • Boundary markers: There are no documented prompt delimiters or specific instructions for the agent to ignore commands within the source files.
  • Capability inventory: The skill uses the Read tool and executes local scripts (bin/setup-retrieve.sh, scripts/retrieve.py, scripts/contextual-prefix.py, scripts/bm25-index.py) using the Bash tool.
  • Sanitization: The skill does not describe any sanitization or validation for the document content it processes.
  • [COMMAND_EXECUTION]: The skill uses the Bash tool to run local scripts for setup, index maintenance, and performing retrieval queries.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 30, 2026, 06:46 PM
Security Audit — agent-trust-hub — wiki-retrieve