The Agent Skills Directory

[SAFE]: No malicious patterns or vulnerabilities were identified. The skill demonstrates high security awareness in its implementation.
[INDIRECT_PROMPT_INJECTION]: The skill processes project requirements and source code to generate testing guides, which is an inherent part of its functionality.
Ingestion points: Project documentation (requirements.md, design.md, tasks.md) and product source code files as analyzed in SKILL.md and scripts/dogfood.
Boundary markers: The skill utilizes a structured JSON schema for case data and isolated subagent briefs to maintain context boundaries during the review process.
Capability inventory: File system write access for generated artifacts and local HTTP server binding on the loopback interface.
Sanitization: Content rendered in the HTML guide is sanitized via escapeHtml and escapeAttr functions within shell/guide.html to prevent cross-site scripting.
[COMMAND_EXECUTION]: The skill executes an internal Python CLI tool (scripts/dogfood). This tool is built using only the Python standard library, which significantly reduces the risk of external dependency vulnerabilities.
[PRIVILEGE_ESCALATION]: The process termination feature in the CLI is secured by a token-based verification system. The script confirms the identity of the process on the local port before issuing a SIGTERM, preventing unauthorized process signaling.

write-dogfood