The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to use shell commands, specifically curl, and to generate 'tiny scripts' to probe live API endpoints. This is the core functionality for verifying real-world API behavior as described in the grounding-checklist.md.
[EXTERNAL_DOWNLOADS]: The agent is directed to pull 'primary source' documentation and actual API specifications from external URLs to ground the integration design before writing code.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests untrusted data (external API specs). While the skill emphasizes that 'reality wins' and focuses on live probes, it lacks explicit boundary markers or instructions to ignore potential malicious content embedded within external documents.
Ingestion points: External API specifications and protocol documentation fetched from user-provided or researched URLs (SKILL.md Step 1).
Boundary markers: None detected; the skill does not instruct the agent to use delimiters or ignore instructions found within the specs.
Capability inventory: The skill has access to shell command execution (curl), script generation, and local file writes to the ./skill-outputs/ directory.
Sanitization: No explicit sanitization or validation logic is provided for the fetched documentation content.
[DATA_EXFILTRATION]: While the skill involves network operations, it incorporates safety best practices by explicitly instructing the agent never to hardcode secrets (keys/tokens) in outputs and to reference them via environment variables like $TOKEN.

spec-grounded-design