The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection as it is designed to ingest and process data from external, potentially untrusted sources. Injected instructions could persist in the research artifact and influence future agent actions. \n
Ingestion points: The skill ingests reader-report.md files during Stage 1 and Stage 3, and performs "number grounding" in Stage 3 by opening and reading project source files, execution logs, and benchmark results to verify data. \n
Boundary markers: Absent. The instructions do not mandate the use of delimiters or "ignore" warnings when interpolating content from these external files into the research record. \n
Capability inventory: The skill uses Read, Write, Edit, Glob, and Grep tools (as defined in SKILL.md) to manage the ara/ artifact, providing a mechanism to persist or act upon data found in external files. \n
Sanitization: Absent. While the skill distills conversational prose into telegraphic language, it does not explicitly sanitize, escape, or validate the content gathered from external files before writing it to the logic or trace layers.

research-manager