The Agent Skills Directory

[DYNAMIC_EXECUTION]: The file references/isolated-verification.md contains a complex template for a Bash and Python script. The agent is instructed to populate variables (such as base commits, file paths, and test commands) and execute this script at runtime. This script performs Git operations (worktree, read-tree, apply), creates temporary directories, and uses Python's subprocess module to verify file integrity and run build/test suites in an isolated environment.
[INDIRECT_PROMPT_INJECTION]: The skill's core functions require reading and processing untrusted repository data to determine task scope, verify requirements, and validate implementation outcomes.
Ingestion points: Data enters the agent's context through git status, git diff, git ls-tree, and search tools like ripgrep as described in references/scope-and-sweeps.md and references/isolated-verification.md.
Boundary markers: The instructions do not define specific delimiters or "ignore embedded instructions" warnings for the repository content being analyzed.
Capability inventory: The skill has broad capabilities to execute shell commands, including build tools, test runners (pytest, npm test), infrastructure tools (terraform, docker, kubectl), and Git management commands.
Sanitization: No explicit sanitization or validation of the repository-supplied content is performed before it influences the agent's logic or is used in command arguments.

verification-before-completion