The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions (SKILL.md, Phase 4) include a final step to execute npm run format. This is a standard development task used to maintain consistent file formatting. The skill explicitly requires the agent to ask for user confirmation before executing the command, adhering to safety best practices.
[INDIRECT_PROMPT_INJECTION]: The skill operates by reading existing documentation and source code (within docs/ and packages/ directories) to ensure technical accuracy. Ingesting content from the repository represents a standard attack surface for indirect prompt injection. However, the skill mitigates this by incorporating a mandatory verification phase (Phase 4) which includes accuracy checks, self-review, and link validation, ensuring that the agent remains focused on the documentation task.
[DATA_EXPOSURE]: While the skill accesses project files and source code, its activities are strictly scoped to the repository's documentation needs. It does not attempt to access sensitive system files (e.g., SSH keys, environment variables) or perform unauthorized network exfiltration.

docs-writer