llm-wiki-capture
Pass
Audited by Gen Agent Trust Hub on Oct 11, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests potentially untrusted data from external sources and session history. Ingestion points: 'references/source-ingest.md' processes external URLs and materials; 'references/session-capture.md' utilizes session history as evidence. Boundary markers: 'references/provenance-contract.md' explicitly defines source content as untrusted evidence and instructs that embedded commands must not influence the agent's workflow. Capability inventory: The skill performs file system writes, Git management (commit, push), and local script execution. Sanitization: 'references/session-capture.md' mandates the redaction of credentials and sensitive personal information.
- [DYNAMIC_EXECUTION]: The skill generates and executes a local maintenance script to verify the wiki's integrity. Evidence: 'references/bootstrap.md' describes the creation of 'scripts/wiki-maintenance', a zero-dependency script executed to validate the schema and local links.
- [COMMAND_EXECUTION]: The skill executes system-level Git commands for repository management. Evidence: 'references/integration-contract.md' and 'references/bootstrap.md' describe the use of git for initialization, status checks, committing, and pushing, all of which require explicit user authorization.
Audit Metadata