The Agent Skills Directory

[SAFE]: No malicious patterns, prompt injections, or security vulnerabilities were identified in the skill instructions, metadata, or reference materials. The skill strictly adheres to a self-contained policy, prioritizing local technical documentation.
[COMMAND_EXECUTION]: The skill guides the agent to perform standard development operations, including environment preflight checks, local package installation (pip install -e .), project builds (ninja, build.sh), and unit/integration testing (pytest). These commands are appropriately scoped to a local development workflow.
[INDIRECT_PROMPT_INJECTION]:
Ingestion points: The skill processes user-supplied text for Goal, Non-goal, and Acceptance criteria during the intake phase in SKILL.md.
Boundary markers: A strict "Working Contract" and structured "Requirement Intake Template" are used to scope the agent's actions and minimize the risk of executing unintended instructions.
Capability inventory: The skill has the capability to execute shell commands for repository searching (rg), building, and testing, as well as running local Python scripts for kernel verification.
Sanitization: The skill relies on a mandatory validation matrix and reproducible command outcomes to verify that generated code and artifacts meet the user's defined acceptance criteria.

tle-developer-flagos