The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process potentially untrusted data from multiple sources including user-provided existing resumes, portfolios, and Job Descriptions (JDs) fetched from the web.
Ingestion points: Direct user input, existing materials copied to the _workspace/ directory, and external content retrieved via WebSearch and WebFetch tools.
Boundary markers: The instructions do not specify explicit delimiters or "ignore embedded instructions" warnings for the agent when processing user-provided career materials.
Capability inventory: The skill uses a team of five agents with the ability to write files to the local filesystem (_workspace/), perform network operations via search/fetch tools, and communicate via SendMessage.
Sanitization: While the skill includes guidelines for abstracting confidential details, it lacks specific sanitization or validation protocols to prevent malicious instructions hidden within processed resumes or JDs from influencing agent behavior.
[COMMAND_EXECUTION]: The skill workflow requires the agent to perform filesystem operations to manage project deliverables.
Evidence: Phase 1 instructions mandate the creation of a _workspace/ directory in the project root and the saving of structured data to _workspace/00_input.md.

personal-branding