The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill generates and executes platform-specific launcher scripts (.sh for POSIX and .ps1 for PowerShell) based on observed repository structure. Additionally, it implements a mechanism to load and follow repository-local instruction extensions from .ai/skills/om-integration-tests/SKILL.md. While the skill includes instructions to ignore safety-relaxing directives, this creates a path for untrusted content within a repository to influence the agent's execution logic.
[INDIRECT_PROMPT_INJECTION]: The skill processes a significant amount of untrusted external data, including PR bodies, issue descriptions, and repository-specific documentation, creating a large surface for indirect prompt injection.
Ingestion points: Reads tracker content (PRs/Issues) and local project instructions like AGENTS.md or CLAUDE.md as defined in references/agentic-setup.md.
Boundary markers: Includes a dedicated 'Untrusted content boundary' section in references/agentic-setup.md that instructs the agent to treat these sources as data only and ignore embedded directives.
Capability inventory: The skill is capable of executing arbitrary shell commands, repository build scripts, and test runners.
Sanitization: Implements regex validation (^[A-Za-z0-9._/-]+$) for external values before they are used in shell or path interpolation.
[COMMAND_EXECUTION]: The skill is designed to discover and execute various commands found within the repository, such as Makefile targets, package.json scripts, and native environment launchers. This capability is inherent to its purpose as a test runner but remains a high-privilege operation when interacting with untrusted repositories.

om-integration-tests