The Agent Skills Directory

[SAFE]: The skill provides a valid scientific drafting workflow that binds evidence records to paper subsections based on relevance and user-defined constraints.
[SAFE]: All file read and write operations are strictly restricted to the local workspace path provided by the user, with no indications of unauthorized path traversal.
[SAFE]: The skill does not perform network operations, use hardcoded credentials, or employ any form of code obfuscation.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection as it processes untrusted data from external sources.
Ingestion points: outline/subsection_briefs.jsonl, outline/mapping.tsv, and papers/evidence_bank.jsonl.
Boundary markers: Absent.
Capability inventory: Reading and writing local JSONL and Markdown files within the workspace using deterministic logic.
Sanitization: The skill processes inputs as data for scoring and tokenization rather than as executable instructions, effectively mitigating the risk of injection affecting script logic.

evidence-binder