Living collection
Productivity and focus | Skills for Claude Code, Codex and Coding Agents by Task
25 resourcesOrganized by use
Productivity and focus
Productivity and focus
3 resourcesFull profiles and sources
Ponytail
Skill
Keep coding-agent context lean
Applies discipline to context and generated code: retain what matters and prune unnecessary work.
Best for: Reducing avoidable code and context in coding-agent sessions.
Why it belongs here. Pushes the agent to solve the actual request with less unnecessary code and context.
Its limit. Shorter output is not automatically better. Missing requirements or useful explanation can cost more in rework.
It asks the agent to do the job with fewer extras. Check that the finished change still covers what you asked for.
Compatibility and setupRequirements and agent differences
Setup. Follow the repository's installation instructions for the agent you use; do not infer support from a generic skill label.
Real cost and licenseResource, agent and extra costs
Practical cost. The useful comparison is total work, including corrections, rather than a token claim taken from a project description.
Example and alternativeWhen it helps and what else to use
Useful route. Try it on a bounded coding task and compare the result, review effort and corrections with the normal workflow.
Sources and reviewWhat is documented and what was tested
Evidence. This entry uses the existing Marketing Ultra review and the project's official source. Author documentation is not described as a hands-on test by Dani.
Resource detailsEditorial profile and links
I have ADHD
Skill
Action-first answers from a coding agent
Reformats agent replies around the next action and numbered steps, with less preamble and filler.
Best for: Coding sessions that benefit from direct instructions and numbered steps.
Why it belongs here. Makes the next action easier to find when a long answer would add friction.
Its limit. A shorter reply can omit context that matters for a difficult decision. Ask for the missing reasoning when the stakes require it.
It changes how the agent talks to you: action first, then the steps. It does not change the underlying model's ability.
Compatibility and setupRequirements and agent differences
Setup. The repository defines the actual installation route; check it before assuming the same setup works in another agent.
Real cost and licenseResource, agent and extra costs
Practical cost. There is no proven productivity gain for every user. Evaluate the reading and correction effort in your own workflow.
Example and alternativeWhen it helps and what else to use
Useful route. Use it for routine coding steps when direct, numbered guidance helps you keep your place.
Sources and reviewWhat is documented and what was tested
Evidence. This entry uses the existing Marketing Ultra review and the project's official source. Author documentation is not described as a hands-on test by Dani.
Resource detailsEditorial profile and links
Claudex Loop
Resource
Combines planning, review and code inspection across Claude Code and Codex, with decisions recorded in the project.
Why it belongs here. Combines planning, review and code inspection across Claude Code and Codex, with decisions recorded in the project.
Its limit. Adds time and usage. Agreement between two models does not prove that a solution works.
One proposes a solution and the other looks for holes. You decide which findings matter, and tests check the result.
Compatibility and setupRequirements and agent differences
Details. The complete workflow needs authenticated Claude Code and Codex CLIs and Python 3.10 or later. Plugin commands use the /claudex-loop: prefix.
Real cost and licenseResource, agent and extra costs
Details. The MIT-licensed code costs €0. Review rounds use both provider accounts and their usage limits or billing.
Example and alternativeWhen it helps and what else to use
Details. Useful for reviewing changes with several decisions. For a small edit, a one-off review avoids running the whole workflow.
Sources and reviewWhat is documented and what was tested
Details. Checked against the official repository on 21 September 2026 and the published review. No independent productivity or cost comparison.