agent-spec-tool-first
Pass
Audited by Gen Agent Trust Hub on Apr 9, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill guides the agent to verify the presence of the 'agent-spec' CLI and suggests installation via 'cargo install agent-spec' if necessary. This targets crates.io, the official package registry for Rust, which is a well-known and trusted service.\n- [COMMAND_EXECUTION]: The workflow relies on executing 'agent-spec' commands for contract management and code verification. Commands like 'lifecycle' and 'guard' execute local project tests to verify that code changes adhere to the provided specifications.\n- [PROMPT_INJECTION]: The skill ingests untrusted data from '.spec' files and source code to drive agent behavior. It mitigates indirect injection risks through a mandatory evidence chain: ingestion occurs via spec files; boundary markers are established using structured contract sections (Intent, Decisions, Boundaries); capabilities are restricted to the 'agent-spec' toolset; and the 'lint' layer provides automated sanitization and quality checks before implementation.
Audit Metadata