write-spec
Pass
Audited by Gen Agent Trust Hub on Sep 26, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to research external sources such as official documentation, source repositories, and papers to resolve unknowns in feature specifications. This creates an ingestion surface for untrusted data that could contain malicious instructions designed to influence the agent's planning.
- Ingestion points: Online research sources and local repository inspection (SKILL.md, First Principles 3).
- Boundary markers: Incorporates human review checkpoints, parity gates for behavior validation, and a synthesis phase to reconcile conflicting drafts.
- Capability inventory: File creation within the specs directory, spawning of subagents for parallel drafting, and the execution of reproduction spikes.
- Sanitization: Relies on the agent synthesizing multiple perspectives and performing a scrollback audit rather than technical input filtering.
- [DYNAMIC_EXECUTION]: The workflow mandates the creation of reproduction spikes when research identifies a reference implementation or unfamiliar technique. This involves the generation and execution of code artifacts to validate architectural assumptions before they are committed to the project (SKILL.md, Workflow 2).
Audit Metadata