The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from the BKD REST API (such as issue logs and descriptions), creating a potential surface for indirect prompt injection.
Ingestion points: Data enters the agent's context through API responses fetched via curl in SKILL.md and reference files like references/rest-api.md.
Boundary markers: The skill uses jq to parse JSON and check for success envelopes, which helps maintain data structure integrity.
Capability inventory: The skill possesses capabilities for network communication (curl) and filesystem/repository operations (git).
Sanitization: The skill mandates the 'never-inline rule' (Rule 10 in SKILL.md), requiring that free-form text be written to temporary files and processed with jq for binary data transfer. This significantly reduces the risk of malicious content being executed as shell commands or mangling JSON payloads during subsequent API interactions.

bkd