build
Pass
Audited by Gen Agent Trust Hub on Oct 19, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes instructions from external markdown files located in
.specs/plans/. - Ingestion points: Reads plan content from
.specs/plans/<slug>.md(SKILL.md, Workflow Step 1). - Boundary markers: None. The agent is instructed to directly follow architectural decisions and implementation items found in the plan.
- Capability inventory: File writing, local command execution (tests/lints), and git commits (Workflow Steps 5, 6, 8).
- Sanitization: None. The skill implicitly trusts the content of the plan files.
- [COMMAND_EXECUTION]: The skill automatically detects and executes project-defined scripts for testing and linting.
- The toolchain detection (Workflow Step 6) triggers commands like
npm test,pytest,go test, andcargo testbased on the presence of manifest files (package.json,pyproject.toml, etc.). - [DYNAMIC_EXECUTION]: The workflow involves implementing code changes and immediately running them via the project's test suite.
- Workflow Step 5 involves the agent implementing logic and tests, which are then executed in Workflow Step 6. This represents a loop of dynamic code generation and execution.
Audit Metadata