The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process documentation from the docs/lore/ directory to guide feature planning and bug fixes. This creates a surface for indirect prompt injection if the documentation files contain malicious instructions.
Ingestion points: The skill reads multiple files from the local repository, including docs/lore/README.md, pitfalls.md, business-rules.md, api-map.md, CONTEXT.md, and various glossary or term files.
Boundary markers: The instructions command the agent to "Brief, don't dump" and produce a "SHORT brief," which provides structural constraints but no explicit security delimiters or sanitization instructions for the ingested content.
Capability inventory: The skill utilizes file read capabilities (via grep and standard reading), file write capabilities for logging feedback to .lore-feedback.jsonl, and file modification capabilities for updating documentation entries and the .gitignore file.
Sanitization: There is no evidence of content sanitization or validation for the data retrieved from the documentation files.
[COMMAND_EXECUTION]: The skill instructs the agent to execute shell-level search operations to locate relevant documentation.
Evidence: The procedure explicitly requires the agent to "grep docs/lore/ for code: references" and "grep its term headings" to facilitate lookups.

lore-consult