The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted text from external influencer agreements, which creates a potential surface for indirect prompt injection attacks.
Ingestion points: The skill is designed to read "any pasted incoming agreement" and "agency paper" provided as user input during the contract review process in SKILL.md.
Boundary markers: The instructions do not define the use of boundary delimiters or specific "ignore embedded instructions" warnings when the agent processes the untrusted contract text.
Capability inventory: The skill possesses the capability to write drafted agreements to the memory/ directory and execute local operations by invoking registry-events.py with data arguments in SKILL.md.
Sanitization: There is no mention of sanitizing, escaping, or validating the ingested contract terms before they are used to generate memos or propose updates to the creator registry.
[COMMAND_EXECUTION]: The skill performs structured operations that involve local script execution to maintain state.
Evidence: The skill instructions in SKILL.md require the agent to submit signed terms (such as usage-rights and rates) as a one-line update to memory/events/creators.ndjson by invoking an operation: propose request to registry-events.py.

contract-helper