doc-maintenance
Pass
Audited by Gen Agent Trust Hub on Sep 17, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The
scripts/doc_audit.pyscript executes localgitcommands to identify the project root and calculate code churn for staleness detection. These calls use argument lists to avoid shell injection and are constrained to retrieving repository metadata. - [INDIRECT_PROMPT_INJECTION]: The skill processes documentation files and project source code through multiple LLM subagents, creating a surface for potential indirect prompt injection.
- Ingestion points: Documentation markdown files in
docs/andmanual/along with codebase source files used for coverage and freshness audits. - Boundary markers: None identified; subagent prompt templates in
references/agent-dispatch.mddo not include explicit delimiters or safety instructions to ignore embedded commands. - Capability inventory: The skill allows for file modification and creation via remediation agents, local subprocess execution in
doc_audit.py, and subtask dispatch via the Task tool. - Sanitization: No explicit sanitization or validation of the codebase content is prescribed before it is processed by the agents.
Audit Metadata