Audits codebases and generates self-contained implementation plans for executor agents.
npx skills add https://github.com/usmangurowa/socialmocks --skill improve-usmangurowa
Please help me install this Agent Skill. Skill: improve Source: https://github.com/usmangurowa/socialmocks/tree/main/.agents/skills/improve Command: npx skills add https://github.com/usmangurowa/socialmocks --skill improve-usmangurowa
plans/ with inlined code excerpts, exact verification commands, scope boundaries, STOP conditions, drift checks, and a status index, so an executor with zero prior context can succeed. - Execution loop: Variants like execute <plan> dispatch an isolated-worktree executor subagent and review its diff like a tech lead, while reconcile keeps the plan backlog current and --issues publishes plans as GitHub issues. - Use Case: Point it at a TypeScript monorepo and ask for a security-focused audit; it maps the repo, verifies each finding against the source, presents a leverage-ranked table, and writes plans another agent can implement and test independently. ## Quick Start Ask the assistant to audit this repository with the improve skill and turn the top findings into executable plans.High-intent search queries and answers about installing and using this skill.
Invoke the skill on the repository; it runs recon to map languages, frameworks, and verification commands, audits across nine categories, vets each finding against the actual code, then writes self-contained plan files under plans/ that an executor agent can follow without any prior context.
It covers correctness bugs, security issues, performance problems, test coverage gaps, tech debt and architecture, dependency and migration risks, developer experience and tooling, documentation, and product direction. Every finding requires file:line evidence, impact, effort, risk, and confidence ratings.
No. The advisor is strictly read-only on source code and may only create files under plans/ or advisor-plans/. Code changes happen only through a separate executor subagent working in an isolated git worktree, whose diff the advisor reviews without ever merging or committing.
Add quick or deep anywhere in the invocation; standard is the default. Quick covers hotspots with top high-confidence findings, deep audits the whole repo with up to eight parallel subagents, and a focus argument like security or perf limits the audit to one category.
The advisor re-runs every done criterion in the worktree, checks the diff against the plan's in-scope file list, reads the new tests, and renders an APPROVE, REVISE, or BLOCK verdict. Merging remains the user's decision; the advisor never pushes or commits to the user's branch.
Skip it for trivial single-file fixes you can make directly, or when the repository lacks any working verification command and you first need a basic build or test baseline. The skill itself flags a missing verification baseline as the first finding in that situation.
Fei2-LabsCompress shell, git, and other command output to reduce token usage.
TyrizxMerge a topic branch into the base branch and resolve conflicts in the main worktree.
krishnakanthb13Automate end-to-end verification of Spring Boot projects across build, tests, and security scans.
telefrekImplement small code changes with TDD and minimal setup.