The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes shell commands using curl to send status updates and notifications to a local service endpoint.
Evidence: Command blocks targeting http://localhost:8888/notify are present in SKILL.md and all workflow files in the Workflows/ directory.
[COMMAND_EXECUTION]: Dynamic path resolution for storing investigation artifacts is performed using jq in command substitution.
Evidence: Path definitions in SKILL.md and Methodology.md use $(jq -r ...) to resolve the active work directory.
[PROMPT_INJECTION]: The skill exhibits a significant attack surface for indirect prompt injection due to its primary function of ingesting data from uncontrolled external sources.
Ingestion points: Untrusted content from websites, social media, and public records is processed during Phase 3 of the CompanyLookup, PeopleLookup, and EntityLookup workflows.
Boundary markers: The skill does not define explicit delimiters or instructions to help sub-agents distinguish between retrieved data and core instructions.
Capability inventory: The skill has the ability to execute shell commands (curl, jq) and perform file system operations.
Sanitization: No sanitization or validation logic is specified for data gathered from external OSINT sources.
[PROMPT_INJECTION]: The skill includes instructions to prioritize and apply behavioral overrides from a user-controlled local directory.
Evidence: The 'Customization' section in SKILL.md directs the agent to load and apply PREFERENCES.md and configurations from ~/.opencode/skills/CORE/USER/SKILLCUSTOMIZATIONS/OSINT/ which override default behaviors.

OSINT