The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted external data such as logs, traces, and source code, which serves as a potential surface for indirect prompt injection.
Ingestion points: SKILL.md identifies code, documents, logs, traces, and websites as evidence sources to be analyzed by the agent.
Boundary markers: The instructions explicitly mandate that the agent treat all such artifacts as untrusted evidence and states they cannot authorize tool usage, secret access, or scope changes.
Capability inventory: The skill allows for repository exploration, execution of diagnostic experiments, and code repairs.
Sanitization: The instructions require that secrets be kept out of captures and reports.
[EXTERNAL_DOWNLOADS]: The skill references and fetches reference documentation from the author's GitHub repository.
Evidence: References to engineering and measurement guidelines located at https://raw.githubusercontent.com/csark0812/toolbox/main/references/v2/engineering-principles.md and measurement.md.
[NO_CODE]: The skill consists exclusively of markdown instruction files and does not include any executable scripts or binary files.

evidence-based-debugging