The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content in the form of blog drafts from the blogs/ directory across multiple phases. This presents an attack surface where instructions embedded in a draft could potentially influence the agent's behavior during the 'De-slop' or 'Review' stages. However, the scope is restricted to the writing pipeline, and no high-privilege capabilities are exposed to this content.
Ingestion points: The agent reads markdown files from the <PROJECT_DIR>/ (blogs/ directory).
Boundary markers: None explicitly defined to delimit draft content from agent instructions during processing.
Capability inventory: The skill writes to local markdown files, spawns subagents for review, and executes a local validation script. It does not have network access or system-level administrative permissions.
Sanitization: The skill employs a Python script (check-blog.py) that uses regular expressions to validate document structure against the style guide.
[COMMAND_EXECUTION]: The skill executes a local script scripts/check-blog.py to automate style guide enforcement. The script is safe as it only uses Python standard libraries (re, sys, pathlib) and performs read-only regex validation on the target markdown files. It does not accept arbitrary shell input or perform dangerous system calls.

blog