The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes data from four independent research sources, creating a surface where malicious instructions in source reports could theoretically influence the final synthesis.
Ingestion points: Technical reports are received from the api-state, ecosystem-research, impact-measurement, and codebase-auditor research angles (SKILL.md, Step 1).
Boundary markers: The skill uses markdown headers to separate angle outputs in the final synthesized note but does not provide specific instructions to the agent to treat ingested data as untrusted or to ignore embedded instructions.
Capability inventory: The skill is strictly limited to text synthesis and conflict resolution; no subprocess execution, file-writing, or network tools are invoked.
Sanitization: The instructions prioritize logical consistency and citation checks but do not define specific sanitization or escaping protocols for the content being processed.

research-note