forge
Pass
Audited by Gen Agent Trust Hub on Sep 12, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes shell commands for repository management and verification. Specifically, it uses
git addandgit committo finalize changes and runs external tools for typechecking, linting, and testing. - [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and act upon potentially untrusted data from external sources.
- Ingestion points: Phase 1 involves ingesting "input (ticket, spec, changeset, or user prompt)" to stage modifications.
- Boundary markers: While the skill defines "Operating Invariants" like "Scope Discipline" to restrict actions to specific criteria, it lacks explicit instructions for the agent to ignore or delimit embedded instructions within the ingested data.
- Capability inventory: The skill has extensive file system capabilities (Create, Move, Update, Delete) and shell execution capabilities (Git, testing tools).
- Sanitization: There are no mentioned mechanisms for sanitizing or validating the contents of the tickets or specifications before they are used to drive the implementation phase.
Audit Metadata