The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts/lookup.js script uses node:child_process.spawnSync to execute a local binary file named s2-docs.js. The arguments provided by the AI agent are passed to this binary. This is standard behavior for a CLI wrapper and uses a safe execution method (array-based arguments) that avoids shell injection vulnerabilities.
[PROMPT_INJECTION]: The skill is designed to ingest and process documentation content from the s2-docs binary. This introduces a surface for indirect prompt injection if the source documentation were to contain adversarial instructions.
Ingestion points: Output from the s2-docs.js binary is read into the script (result.stdout).
Boundary markers: The script uses clear delimiters (=== S2 DOCS: ... ===) to separate tool output when the --cursor flag is used, which helps the agent distinguish between data and instructions.
Capability inventory: The skill has permissions to execute commands via the allowed-tools configuration in SKILL.md.
Sanitization: The script includes basic sanitization by truncating documentation output to 3000 characters to prevent context exhaustion attacks.

s2-docs