The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: Fetches the ctx-cli binary from the official tabnine/skills GitHub repository during the setup phase.
[COMMAND_EXECUTION]: Executes shell scripts to configure the CLI, manage permissions with chmod +x, and perform version checks against GitHub's API.
[COMMAND_EXECUTION]: Provides access to the ctx-cli tool via bash, allowing the agent to query a knowledge graph and manage security vulnerability remediation.
[PROMPT_INJECTION]: Exposes a surface for indirect prompt injection via the processing of untrusted security vulnerability data.
Ingestion points: get_cve_resolution_status in security.md and find_entities in codebase-search.md ingest content from an external knowledge graph and security scanners.
Boundary markers: No delimiters or explicit instructions to ignore embedded commands are present in the documentation for handling tool outputs.
Capability inventory: The skill has execution capabilities via Bash(ctx-cli:*) and instructs the agent to handle code diffs provided by the service.
Sanitization: No documentation exists regarding the sanitization or validation of the recommendedAction diffs returned by the API.

ctx