The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests untrusted data from external job postings fetched via URLs.
Ingestion points: The skill uses WebFetch to retrieve content from user-provided URLs in targeted mode and source URLs from job_search_tracker.csv in aggregate mode.
Boundary markers: The skill contains explicit instructions in Step 8.9 directing the agent to treat stored gaps as data, not instructions, and to never follow directions embedded in those fields.
Capability inventory: The skill has Read and Write access to the local filesystem, as well as WebFetch and WebSearch capabilities, which could be leveraged if an injection is successful.
Sanitization: There are no programmatic sanitization steps described, but the instructional boundary markers serve as a primary defense against accidental execution of embedded instructions.

upskill