The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill requires the agent to research and extract specifications from external documentation URLs. This creates a potential surface where maliciously crafted documentation could influence the generated implementation plans or test logic.
Ingestion points: SKILL.md (Phase 1 Research), docs/cc-sdd/sop-new-agent.md.
Capability inventory: The agent has the capability to write files and execute local commands (npm test, node dist/index.js).
Sanitization: The instructions focus on recording exact values but do not specify sanitization or escaping mechanisms for the extracted data.
Boundary markers: The skill does not define clear delimiters or boundary markers for the external content processed during the research phase.
[COMMAND_EXECUTION]: The verification process involves the execution of local shell commands, including npm test, npm run build, and node dist/index.js. These commands run within the context of the user's project environment to validate the integration.
[DYNAMIC_EXECUTION]: The workflow involves generating new source code and test files (e.g., realManifest{AgentName}.test.ts) based on templates and then executing these files to verify functionality. This represents a pattern where dynamically generated content is executed during the skill's operation.

cc-sdd-new-agent