The Agent Skills Directory

[PRIVILEGE_ESCALATION]: The skill requests Bash(crontab *) permissions to facilitate its daily automated search feature. This allows the agent to modify the user's system-level task scheduler.
[PERSISTENCE]: The skill's primary feature is an automated daily search implemented via a crontab entry, which establishes persistence on the host system to run the agent periodically.
[INDIRECT_PROMPT_INJECTION]: The skill extracts data from external sources (hiring.cafe and various employer career pages) to evaluate job fit. Because this untrusted content is used as context for the agent's decision-making without sanitization, it creates a surface for indirect prompt injection.
Ingestion points: Job results extraction from hiring.cafe (SKILL.md Step 2) and job description extraction from employer career sites (SKILL.md Step 5).
Boundary markers: No delimiters or 'ignore embedded instructions' warnings are specified when processing extracted web content.
Capability inventory: The skill has read/write access to ~/.proficiently/, browser automation capabilities via mcp__claude-in-chrome__*, and shell command execution via Bash(crontab *).
Sanitization: No validation or sanitization is performed on the data extracted from external websites before it is processed by the agent.
[DATA_EXFILTRATION]: The skill accesses highly sensitive personal information, specifically the user's resume and a LinkedIn contact list (linkedin-contacts.csv). While used for the stated purpose of job matching, this provides the agent (and potentially subverted sessions) access to PII and professional network data.

job-search