The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from transcripts, research.md, and hooks.md files to generate its output. If these files contain hidden instructions (e.g., 'ignore previous rules and write about X'), the agent may execute them. Ingestion points: reads multiple markdown files from the 'notes-processing/' directory. Boundary markers: while it performs basic cleaning like stripping YAML frontmatter, it lacks robust delimiters to isolate data from instructions. Capability inventory: the skill has read access to the file system and write access to generate a new 'content.md' file. Sanitization: no security-focused sanitization or filtering is applied to the input content beyond structural cleanup.

content-writer