The Agent Skills Directory

[COMMAND_EXECUTION]: The skill coordinates software debugging and repair by executing build commands and test cases on local source code.
Evidence: SKILL.md instructions to "Rebuild the failed target" and "Run the original failing case".
[INDIRECT_PROMPT_INJECTION]: The skill ingests external data such as failure logs, status reports, and technical documentation, creating a potential surface for indirect instruction injection.
Ingestion points: Ingests failure logs, "Doctor" status reports, and failure evidence files as specified in SKILL.md.
Boundary markers: No explicit delimiters or instructions to ignore embedded commands within logs are defined.
Capability inventory: Uses shell execution for rebuilding targets and running diagnostic test cases, along with file searching via rg (ripgrep).
Sanitization: The skill mitigates risks by mandating SHA-256 integrity verification of the technical basis (SDK_REFERENCES) against a locked hash before processing any documentation.

pva-debug