The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill installs the official benchling-sdk package version 1.25.0 from PyPI. This is a trusted dependency from a well-known vendor, and version pinning is used to ensure environment stability.
[DATA_EXPOSURE]: The skill utilizes environment variables for Benchling authentication, including BENCHLING_API_KEY and BENCHLING_CLIENT_SECRET. It provides clear instructions against hardcoding credentials and emphasizes reading only scoped environment variables.
[INDIRECT_PROMPT_INJECTION]: The skill ingests laboratory data through the Benchling SDK and EventBridge payloads. Evidence Chain: 1. Ingestion points: benchling.dna_sequences.list(), benchling.entries.get_by_id(), and Lambda event details. 2. Boundary markers: No explicit prompt interpolation shown in snippets. 3. Capability inventory: Bash tool and file/network access. 4. Sanitization: Not explicitly implemented. The implementation demonstrates safe data handling focused on programmatic processing.
[DYNAMIC_EXECUTION]: No evidence of runtime code compilation, injection, or unsafe deserialization was found in the scripts or documentation.
[METADATA_POISONING]: The skill's author is listed as "K-Dense Inc." in the metadata, which differs from the vendor identifier "kinandigital". This minor inconsistency does not appear to facilitate deceptive behavior or misjudgment of the skill's safety.

benchling-integration