The Agent Skills Directory

[COMMAND_EXECUTION]: The skill directs the agent to execute local shell commands, specifically git for maintaining a versioned trail of the bibliography progress and grep for verifying that citation extraction from the manuscript was successful.
[REMOTE_CODE_EXECUTION]: The skill includes Python code snippets (located in phase1-extraction.md) intended for the agent to execute locally to perform normalization, deduplication, and parsing of citation data.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection due to its processing of untrusted manuscript content. Ingestion points: User-provided manuscript file (manuscript.md) and the local bibliography library (references.bib). Boundary markers: No explicit markers or 'ignore' instructions are used when reading external content. Capability inventory: Subprocess execution (git, grep), file writing (bibliography/bibliography.md), and Python code execution. Sanitization: The skill relies on specific regular expressions to extract citations, which provides a level of structural validation but does not sanitize the input against sophisticated instructions embedded within text that matches the citation pattern.

bibliography-builder