The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted conversation history to generate a handoff document, creating a potential surface for indirect prompt injection. Ingestion points: Processes the active conversation history (SKILL.md). Boundary markers: No specific delimiters are mandated for the summary output. Capability inventory: File writing to the system's temporary directory. No shell execution or network exfiltration capabilities are present across the provided files. Sanitization: The skill contains explicit instructions to redact any sensitive information, such as API keys, passwords, or personally identifiable information, which reduces data exposure risks.\n- [EXTERNAL_DOWNLOADS]: The documentation references installation commands and source repositories for provenance and installation context. These point to a public repository of a known developer and do not involve execution of unverified remote code during skill runtime. Evidence: References to the source code repository and npx installation commands in SKILL.md and SOURCE.txt.

pm-handoff