The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill establishes an integration pattern that ingests data from external API sources, which represents a potential surface for indirect prompt injection attacks.
Ingestion points: External API response bodies are processed by the APIClient class in SKILL.md using response.json().
Boundary markers: The instructions lack explicit guidance on using protective delimiters or 'ignore' instructions when the agent processes data fetched from external endpoints.
Capability inventory: The skill template includes full network request capabilities via the httpx library and logging functionality.
Sanitization: While the skill suggests redacting sensitive data in logs and using schema validation, it does not provide patterns for sanitizing payloads against embedded malicious instructions.
[COMMAND_EXECUTION]: The debugging section provides the agent with specific shell commands to execute for network diagnostics.
Evidence: The connectivity checklist in SKILL.md instructs the agent to run curl -v {base_url}/health to verify API availability.

api-integration-architect