Build

Execute implementation plan tasks with TDD workflow, auto-commit, and phase gates. Use when user says "build it", "start building", "execute plan", "implemen...

This skill will read and modify files in your repository (plan.md, workflow docs, stack yaml, progress logs), run shell commands (make, npm/pnpm, linters/tests), and may install project git hooks. It requests no secrets, but it may invoke MCP tools (code search/web search) that could send queries or metadata to external services — decide if you want your project context exposed. Before running: (1) review the referenced files (docs/plan/, docs/workflow.md, templates/stacks/*, .solo/pipelines/progress.md); (2) ensure no secrets or CI tokens are committed; (3) verify any commands the skill will run (there is a possible typo like 'uv run pre-commit install'); and (4) consider running on a branch or cloned copy so commits/hooks are reversible. If you want tighter control, restrict the skill to projects you trust or run it in an isolated environment.

Type: OpenClaw Skill Name: solo-build Version: 2.2.1 The skill is classified as suspicious due to its extensive use of powerful tools like `Bash`, `Write`, `Edit`, and `web_search` as defined in `SKILL.md`. While the instructions consistently guide the agent towards legitimate software development, testing, and verification tasks (e.g., running linters, compilers, test suites, git commands, installing dev tools), the inherent capability to execute arbitrary shell commands and modify files presents a significant attack surface. A compromised agent or malicious project configuration (e.g., in `Makefile`, `docs/workflow.md`, or `stack.yaml`) could exploit these capabilities for unauthorized actions, even though the skill's own instructions do not demonstrate malicious intent.