The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data in the form of call transcripts via the blockrun_voice_status tool. This content is generated by external parties on a phone call and could contain adversarial instructions designed to influence the agent's behavior during transcript summarization.
Ingestion points: Call transcripts returned by the blockrun_voice_status endpoint described in SKILL.md.
Boundary markers: Absent; the tool definitions do not specify the use of delimiters or instructions to ignore embedded commands within the transcript data.
Capability inventory: The skill is limited to making network requests to a local proxy gateway (localhost:8402) for telephony services.
Sanitization: Absent; the skill does not describe any mechanisms for filtering or sanitizing the transcript text before it is presented to the agent.
[NO_CODE]: The skill consists entirely of tool definitions and instructional markdown without any accompanying executable script files (e.g., .py, .js, .sh).

phone