The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses local project management and developer CLI tools, such as lev and bd, to manage the lifecycle of plans and task entities.
[DYNAMIC_EXECUTION]: The skill implements a mechanism for defining and running automated checks. It generates requirement rows containing shell commands (via the check: property) and executes them to verify that specific conditions are met before a task is finalized.
[INDIRECT_PROMPT_INJECTION]: The skill ingests intent and requirements from external sources (captures, designs, and tracker URLs), creating a surface for indirect instructions to influence the generated plan or task commands.
Ingestion points: Data from external design documents, capture sessions, and issue trackers is resolved and loaded during the bind_source and rehydrate_fidelity workflow steps.
Boundary markers: The skill freezes source intent into a structured "transition packet" and organizes requirement sentences into discrete YAML rows to maintain separation.
Capability inventory: The skill is capable of writing plans and task packets to the local .lev/ directory and executing shell commands specified in requirement checks.
Sanitization: The workflow includes deterministic structural validation (e.g., lev task validate --json) and a semantic readiness review that assesses intent fidelity and outcome demonstrability.

lev-plan