The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a local binary tool named newsjack for coverage tracking and state management.
Evidence includes shell commands: newsjack --version, newsjack coverage status <slug>, newsjack coverage check <slug> --input candidates.json, and newsjack coverage record <slug> --input decisions.json.
The instructions explicitly direct the agent to check for the presence of this binary in standard and custom paths (~/.newsjack/bin/newsjack).
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted data from external sources, creating an attack surface for indirect prompt injection.
Ingestion points: Data enters the agent's context through the news-search tool output, snippets, and full article content when snippets are insufficient (found in SKILL.md under Workflow steps 2 and 4).
Boundary markers: There are no explicit instructions to use delimiters or sanitization for the untrusted news content before processing or interpolating it into the agent's reasoning process.
Capability inventory: The skill has capabilities including shell command execution (newsjack), filesystem writes (temporary JSON files and SQLite database), and network access (via news-search or browsing tools).
Sanitization: The skill lacks instructions for sanitizing or escaping article content before the LLM classifies it.

coverage-tracker