The Agent Skills Directory

[SAFE]: The skill operates as a set of text-processing instructions and does not execute arbitrary code or shell commands. It includes a comprehensive Copy Editing Spine and Final QA process that enforces strict source fidelity, preventing the agent from generating content not present in the allowed inputs.
[SAFE]: The skill discovery mechanism for repository context files (e.g., .agents/marketing-context.md) uses a fixed search order and does not accept user-controlled paths, effectively preventing path traversal or unauthorized file access.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a data ingestion surface as it processes user-provided drafts and marketing context files. However, this is mitigated by the Significant Ambiguity and Source Boundary rules which instruct the agent to prioritize source fidelity and ask clarifying questions rather than following potentially malicious embedded instructions. 1. Ingestion points: User-pasted text and marketing context files (.agents/marketing-context.md, .claude/marketing-context.md, marketing-context.md). 2. Boundary markers: The SKILL.md and references/source-boundary.md files provide explicit delimiters and fidelity instructions. 3. Capability inventory: The skill is limited to text analysis and rewriting; no system-level or network capabilities are utilized. 4. Sanitization: Results are validated against the references/final-qa.md checklist to ensure all claims are traceable to verified sources.

a1-copy-editing