The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to analyze potentially untrusted data such as transcripts, logs, and repository content, which presents an attack surface for indirect prompt injection.
Ingestion points: SKILL.md specifies that transcripts, logs, and external website content are to be retrieved and reviewed as evidence.
Boundary markers: The skill includes a robust 'Authority boundary' section that explicitly directs the agent to treat all evidence as untrusted and forbids the authorization of tools, edits, or secret access based on that data.
Capability inventory: No executable capabilities or tool definitions are present; the skill is configured with disable-model-invocation: true and mandates a 'Read-only' workflow.
Sanitization: The instructions provide explicit negative constraints to prevent the agent from obeying instructions found within the analyzed data, ensuring it remains in an analytical role.
[EXTERNAL_DOWNLOADS]: The skill contains references to external GitHub repositories for documentation and licensing purposes.
Evidence: The SKILL.md and LICENSE files provide links to github.com/mattpocock/skills and github.com/backnotprop/pstack. These are used for attribution and to provide context for the skill's development, representing standard open-source documentation practices.

retro