claude-exec
Pass
Audited by Gen Agent Trust Hub on Oct 5, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill facilitates the processing of untrusted data by a sub-agent, creating a potential surface for indirect prompt injection where malicious content in files could influence the sub-agent's actions.\n
- Ingestion points: The skill ingests data from
PLAN.mdand the output ofgit diffcommands.\n - Boundary markers: The delegated prompts lack explicit delimiters or instructions to the sub-agent to ignore embedded commands.\n
- Capability inventory: Sub-agents are granted access to powerful tools such as
Bash,Read,Glob, andGrep.\n - Sanitization: No sanitization or escaping of external data is described before interpolation into the sub-agent prompt.\n- [COMMAND_EXECUTION]: The skill provides instructions for executing the
claudeCLI fallback via shell commands. Although it recommends safety flags like--permission-mode dontAsk, this involves the generation and execution of shell commands with user-defined input.
Audit Metadata