The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill reads bearer authentication tokens from ~/.convex/config.json. This sensitive file contains credentials required for interacting with the Convex cloud platform.\n- [COMMAND_EXECUTION]: The skill executes git status, git worktree list, and gh pr list commands to inventory local work and remote pull request states.\n- [EXTERNAL_DOWNLOADS]: The skill communicates with the official Convex Management API at https://api.convex.dev/v1. Convex is a well-known service, and the network activity is consistent with the skill's management purpose.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes pull request lists and repository files, which can contain untrusted content from external contributors.\n
Ingestion points: Pull request metadata via gh pr list and local documentation files like AGENTS.md.\n
Boundary markers: The skill instructions do not specify delimiting or escaping the output of the CLI tools.\n
Capability inventory: The agent has the ability to delete cloud deployments and execute shell commands based on the analyzed data.\n
Sanitization: No explicit sanitization or validation of pull request names or repository file content is performed before processing.

convex-cleanup