The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses standard CLI tools including git status and grep to manage file state and search for content within the user's project directory. These commands are used legitimately for the stated purpose of identifying target files and ensuring a clean working environment before and after modifications.\n- [PROMPT_INJECTION]: The skill processes untrusted markdown files, representing a surface for indirect prompt injection. The risk is mitigated by the architectural requirement for human review.\n
Ingestion points: Ingests content from content/posts/*.md during scanning and preview phases as described in SKILL.md.\n
Boundary markers: Identifies YAML frontmatter delimited by --- but lacks explicit delimiters for the main body content.\n
Capability inventory: Possesses Bash, Write, Edit, and Task tools, allowing for file modification and command execution across all scripts.\n
Sanitization: No automated content sanitization is performed on the files being read; however, the mandatory 'Preview-Confirm-Apply' workflow provides a human-in-the-loop safety gate that prevents the agent from autonomously following instructions embedded in posts.

batch-editor