The Agent Skills Directory

[DYNAMIC_EXECUTION]: The script scripts/dispatch-task-execution.cjs can execute a project-local script specified in the workspace configuration under execution_routing.resolver.script. This script is dynamically invoked using execFileSync to determine task routing logic.
[COMMAND_EXECUTION]: The skill's primary function is to coordinate and execute commands. It spawns subprocesses for various external AI CLI tools such as claude, codex, cursor-agent, gemini, copilot, and opencode. It also executes several internal helper scripts to manage the task lifecycle.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted Markdown content from task files and interpolates it into prompts sent to external execution harnesses, creating a vulnerability surface.
Ingestion points: scripts/dispatch-task-execution.cjs reads the full content of task Markdown files.
Boundary markers: Absent; the taskPrompt function appends raw task Markdown to the end of its instruction block without using delimiters or instructions to ignore embedded commands.
Capability inventory: The skill possesses significant capabilities, including arbitrary command execution (via harnesses), file system writes (updating status), and network access (update checks).
Sanitization: No filtering, escaping, or validation of the task file content is performed before it is included in the outgoing prompt.
[EXTERNAL_DOWNLOADS]: The script scripts/check-for-updates.cjs performs network requests to the GitHub API (https://api.github.com/repos/e0ipso/strikethroo/releases/latest) to check for newer versions of the Strikethroo tool. This request targets the author's official public repository.

st-execute-task