ia-compound-docs
Pass
Audited by Gen Agent Trust Hub on Jul 16, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands such as
mkdir,grep,echo, andcatfor directory management and file creation. It also executes a bundled bash script,scripts/validate-frontmatter.sh, to ensure YAML metadata conforms to the required schema. - [PROMPT_INJECTION]: The skill presents an indirect prompt injection surface because its primary function is to ingest and process untrusted data from user conversations to generate documentation.
- Ingestion points: Conversation history accessed during the context gathering phase (Step 2 in
references/documentation-process.md). - Boundary markers: None explicitly defined in the instructions to separate extracted user content from the documentation templates.
- Capability inventory: Persistent filesystem writes, directory creation, and command execution capabilities are used to store the ingested data.
- Sanitization: The skill includes filename sanitization (Step 4) and strict YAML enum validation (Step 5), which mitigates some risks related to malicious metadata or paths, but doesn't sanitize the technical resolution content itself.
Audit Metadata