The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill provides instructions for the agent to implement and review webhook handlers that ingest untrusted data from third-party providers, creating an attack surface where malicious payloads could influence agent behavior.
Ingestion points: The skill documentation in SKILL.md includes code patterns for reading raw request bodies and parsing JSON data.
Boundary markers: No specific LLM boundary markers or "ignore embedded instructions" warnings are defined for the processed data, although the skill emphasizes cryptographic verification as a security boundary for the application.
Capability inventory: The skill is configured with access to Read, Grep, Bash, and Edit tools as specified in the allowed-tools metadata in SKILL.md.
Sanitization: The skill promotes best practices for validating external content, such as HMAC-SHA256 signature verification and timing-safe comparisons.
[EXTERNAL_DOWNLOADS]: The skill references an external JSON artifact for evaluation purposes.
Evidence: A link to a comprehension-eval artifact at https://github.com/jacob-balslev/skill-graph/blob/main/examples/evals/webhook-integration.json is provided in SKILL.md.
Context: The referenced repository belongs to the skill's author infrastructure.

webhook-integration