Senior-thinking-skills
A set of 29 skills that make an agent think like a senior engineer before writing code: plans, failure causes and review
- 89 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- The set includes optional hooks and scripts that run alongside agent tool calls, advising only by default.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/senior-thinking-skillsDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The set includes optional hooks and scripts that run alongside agent tool calls, advising only by default.
This is third-party code. Review the repository files before installing.
What it does
The set splits engineering thinking into small skills in place of one long one. The single router, `senior-engineer-mindset`, reads the situation and picks the right disciplines. The other twenty-eight load by description: `clarify-the-real-problem` finds the goal behind a request, `weigh-tradeoffs` compares options, `premortem` works through failure scenarios, `bite-sized-plan` cuts work into two to five minute steps, `plan-on-disk` keeps the plan in files, `root-cause-discipline` leads to the root cause and `surgical-change` keeps out lines unrelated to the request. There are verification and review disciplines: `verify-before-claiming`, `fresh-context-review`, `adversarial-review` and `security-review`. Each skill folder holds a thin `SKILL.md` and files tuned to different Claude and GPT models, and the agent reads the one matching its model. Two skills, `delegate-to-subagents` and `senior-engineer-mindset`, include optional PreToolUse and PostToolUse hooks. By default hooks only advise, and strict mode is switched on through an environment variable.
Who it is for. Developers who ask an agent to design a feature, choose a library or find the cause of a failure and want a considered approach on top of the first idea.
Good fit when
- Before a new feature the agent clarifies the real goal and compares several approaches
- A bug is not obvious and needs a root-cause search backed by evidence
- A large job follows a plan in files that survives `/clear` and context compaction
- A diff is reviewed in fresh context on two axes: spec compliance and code quality
Not a fit when
- A one-line fix or a throwaway prototype, which the router itself says to skip
- The plan is fully specified and the agent only has to type it out
- You need a narrow rule set for one language or framework, while these disciplines are general
Example request
We need to choose between Postgres and SQLite for a new service: clarify the goal first, compare options and record the decisionLimitations
The skill descriptions and model files target specific Claude and GPT generations, and on an unlisted model the author advises the general file. The README has no install command: skills are copied folder by folder into the agent skills directory, whole or one at a time. The `delegate-to-subagents` and `senior-engineer-mindset` hooks only advise by default, and strict blocking is switched on with the environment variable `*_HOOK_STRICT=1`. Several skills adapt ideas from other repositories, so adjust the wording to your own rules.
How to disable. Delete the skill folders from `~/.claude/skills` and remove the PreToolUse and PostToolUse hooks from settings if you wired them in.
FAQ
How do I install the set?
Copy the skill folders into the agent skills directory, for example `~/.claude/skills`. The README allows three routes: install everything, take only the disciplines you need, or fork and reword for your team.
Which skill should I start with?
With `senior-engineer-mindset`, the router that picks disciplines and scales the depth to the task. The other skills also work on their own.
Do the hooks block agent actions?
Not by default, they only warn. Blocking mode is switched on with a variable like `*_HOOK_STRICT=1`.
Do I need the whole set?
No, each discipline stands alone. `root-cause-discipline` is enough for debugging, and `bite-sized-plan` with `plan-on-disk` for plans.
Which models are the skills tuned for?
For the Claude and GPT generations listed in the README, with a file per model in every skill. For an unlisted model the author points to the general `sonnet-5.md` file.
Discussion
No questions or reviews yet.
Related
Skills14 skills that find the gaps in a task before the code: interviews, plans, assumption tests and proof that a bug fix works
A library of 30-plus instruction skills for coding agents: planning, building, testing, hallucination guards and shipping
A set of independent skills against drift in large codebases: hard-cut, root-cause tracing, lean testing and CI right-sizing.
Builder Skills
Skills for agents
A set of small skills from Builder.io for agent work habits: visual plans and diff recaps, docs-first checks, cross-agent review and clean status updates
Nine short skills: tests, review, ADR plans, slides, a wiki, image generation, worktrees and plan implementation.
Over twenty skills for coding agents: MCP and hooks, plan mode, a bug-fix protocol, prompting and a bash safety hook
README in short
The README explains the set structure of a router plus disciplines, lists every skill by work stage, three ways to use it, design principles and idea sources, and then keeps a detailed changelog.