The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses shell scripts and Python one-liners to automate its defensive mechanisms and maintenance tasks:
hooks/circuit-breaker.sh: Uses git stash and JSON-based tracking to stop agents from repeatedly editing the same file.
hooks/dead-mans-switch.sh: Monitors commit frequency and stashes uncommitted work if a timeout is reached.
hooks/watchdog.sh: Runs periodic health checks on the development server and architectural layers using curl and python3.
install.sh: Configures the local environment and merges hook configurations into .claude/settings.json using an inline Python script.
[SAFE]: No malicious patterns such as prompt injection, data exfiltration, or obfuscation were detected. The skill's operations are transparent and strictly focused on improving the reliability and observability of the agent's work within the local repository.
[EXTERNAL_DOWNLOADS]: Templates like init.sh.template contain commented-out examples for standard package managers (npm install, pip install), but the skill does not autonomously download or execute external code from the internet.

harness-engineer