The Agent Skills Directory

[COMMAND_EXECUTION]: The skill describes the orchestration and execution of shell commands through the asc workflow run command. These commands are defined in a repository-local configuration file (.asc/workflow.json) and are executed using bash -o pipefail -c or sh -c as specified in the 'Shell behavior' section.
[PROMPT_INJECTION]: The skill defines a system that reads and executes logic from repository-local files, which presents a surface for indirect prompt injection. If an agent processes a repository containing a malicious .asc/workflow.json file, it could be induced to execute arbitrary shell commands.
Ingestion points: The agent is instructed to read workflow definitions from .asc/workflow.json in the active repository.
Boundary markers: The provided instructions do not include specific delimiters or boundary markers to differentiate the JSON-defined command strings from the agent's internal logic.
Capability inventory: The skill facilitates arbitrary shell execution via the asc tool's workflow runner, which can call both built-in asc commands and generic shell commands.
Sanitization: The skill documents a --dry-run capability which allows for the inspection of commands in stderr before actual execution takes place.

asc-workflow