The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill reads and processes untrusted source code files, which constitutes a surface for indirect prompt injection. This risk is amplified by the skill's capability to execute shell commands (git, npm, curl) and its design for multi-step agent chains where audit findings are passed to a secondary executor. Evidence: Ingestion occurs during the recon and audit stages described in references/audit.md; Capabilities include command execution in references/execute.md. Boundary markers are established in SKILL.md requiring read-only auditing and isolated worktree execution. Sanitization relies on agent logic and human oversight.
[COMMAND_EXECUTION]: The skill's verification protocol requires executing shell commands such as git worktree, npm run build, and npx vitest. It also utilizes curl for local testing of application endpoints to verify security fixes.
[EXTERNAL_DOWNLOADS]: The skill includes references to documentation and optional UI assets hosted at otf-kit.dev. These links are provided by the skill author to assist with remediation of design or infrastructure findings.

ship-ready