The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's core functionality relies on processing untrusted external data from pull request diffs and Git history.
Ingestion points: The agent is instructed to read hunks and commit metadata using git diff and git log (Step 0 and Step 1 in SKILL.md).
Boundary markers: The instructions lack specific guidance on using delimiters or strict isolation prompts to prevent the agent from following instructions embedded within the processed diff content.
Capability inventory: The skill possesses significant capabilities, including the Edit tool to modify repository files and the Bash tool to execute shell commands and post PR comments via the GitHub CLI.
Sanitization: There is no evidence of content sanitization or filtering applied to the diff data before it is interpreted by the agent.
[COMMAND_EXECUTION]: The skill requires a wide range of shell commands to perform its analysis and validation workflow.
Evidence: The agent executes commands such as grep, git, and python3 for repository searches, diff resolution, and running local testing scripts (lint_suite.py, run_suite.py).
Context: These operations are scoped to the project environment and are standard for development-oriented agents, but represent an active command execution surface.
[EXTERNAL_DOWNLOADS]: The skill integrates with external tooling and configuration provided by the vendor and trusted sources.
Evidence: Reference documentation (pr-automation.md) describes utilizing npx to fetch skill bundles from the cekura-ai vendor and using the anthropics/claude-code-action in automation workflows.
Context: These resources are identified as originating from the skill's specific vendor or established trusted organizations.

cekura-bot-test-writer