The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external project files which constitutes an ingestion surface for potential indirect instructions.
Ingestion points: Reads content from .tex and .bib files to identify citation keys and bibliography entries (SKILL.md).
Boundary markers: The skill does not define specific delimiters for the ingested file content.
Capability inventory: Authorized to perform file writes, edits, and execute specific shell commands including uv run for verification (SKILL.md).
Sanitization: The extraction process uses regular expressions to isolate citation keys, and the final state change is protected by an explicit user approval step.
[COMMAND_EXECUTION]: The skill utilizes a restricted set of shell utilities to perform administrative tasks.
Evidence: Uses shasum to detect concurrent file changes and uv run to execute a shared output verification script at the end of the workflow (SKILL.md).

bib-filter