dead-letter
Pass
Audited by Gen Agent Trust Hub on Oct 5, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests failure signals such as error logs, terminal stderr, and subagent exit messages which are inherently untrusted (SKILL.md, Step 1). This content is used to determine recovery decisions and author precondition checks.\n
- Ingestion points: Error logs and terminal output ingested during the triage procedure (SKILL.md, Step 1).\n
- Boundary markers: There are no explicit instructions to use delimiters or ignore-embedded-instruction warnings for the ingested failure data.\n
- Capability inventory: The skill utilizes bash, write_to_file, and view_file tools to execute verification and manage records.\n
- Sanitization: The procedure does not specify sanitization or escaping of the ingested signals before they are used to formulate state verification commands or retry packets.\n- [COMMAND_EXECUTION]: The skill executes shell commands for system state verification and status sweeps via the bash tool.\n
- Evidence: The 'Precondition Check' (SKILL.md, Step 4) involves emitting a command or query to verify system state, and the 'Status Sweep' (references/sweep-protocol.md) uses shell tools to inventory and cluster records.
Audit Metadata