agentsclimarketplace

Coding agent workflow

Skill Amey-Thakur/AI-SKILLS/skills/llm-engineering/coding-agent-workflow

Plug-and-play skills and prompts for every AI coding agent

Install
npx -y skills add Amey-Thakur/AI-SKILLS --skill coding-agent-workflow

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

2 things to look at

  • 19 days oldThe repository was created 19 days ago. New is not bad, but a brand new repository carrying a familiar-sounding name is the shape a typosquat arrives in, and there has been no time for anyone else to find a problem with it.
  • 4 stars4 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Direct and supervise a coding agent (Claude Code, Cursor, Antigravity) so it ships correct, verified work. Use when delegating engineering tasks to an AI agent and you want reliable results, not plausible-looking ones.

SKILL.md

2.9 KB, 623 tokens by cl100k_base, as published. Nobody here has run it

Coding agent workflow

A coding agent is fast and tireless but confidently wrong when under-directed. Getting reliable work out of one is a skill: brief it tightly, keep changes reviewable, insist on verification, and stay the accountable engineer, not a spectator.

Method

  1. Brief with a goal and a done-condition, not a vibe. State exactly what you want, the constraints, and how to know it is complete (tests pass, the behavior works, the output matches). A vague task gets a vague, plausible result. Point the agent at the relevant files and conventions rather than letting it guess (see agent-complete-task for the prompt shape).
  2. Keep changes small and reviewable. One task per run; a focused diff you can actually read beats a sprawling change that touches forty files. Large tasks get decomposed into steps you review between (see agent-task-breakdown, pull-request-size's logic applied to agents).
  3. Make it explore before it edits. The agent should read the code and match its patterns, not impose a generic style. Instruct it to understand first; unfamiliar-codebase mistakes come from skipping this (see agent-explain-codebase).
  4. Demand verification, and check it. Require the agent to run the tests, build, and linter, and to exercise the actual behavior, then report what it ran. Then verify yourself: agents claim success they have not demonstrated. "It should work" is not "it works" (see agentic-loops' verify-before-done rule, verify skill).
  5. Review the diff like any code review. Read what changed and why, not just whether it runs. Watch for scope creep (unasked refactors), silent behavior changes, weakened tests, and confident comments over subtly wrong logic. You are accountable for merging it (see review-my-code, code-review).
  6. Gate the irreversible. Destructive or outward-facing actions (deleting, force-pushing, deploying, sending) get your confirmation, not the agent's autonomy; keep it working on a branch, and keep the changes reversible (see llm-guardrails, automation-guardrails).

Boundaries

  • The agent accelerates the work; it does not transfer accountability. Merging unreviewed agent output because it looked confident is how bad changes ship.
  • Agents excel at well-specified, verifiable tasks and struggle with ambiguous ones; the payoff tracks how well you can state the goal and check the result. Underspecified creative architecture is still yours.
  • This is human-directs-agent supervision; multi-agent orchestration (agents coordinating agents) is a separate discipline with its own costs (see agent-orchestration-antipatterns, multi-agent-workflow).

Gives 0 of the 12 instructions most automation workflows skills give in 623 tokens

Counted across 745 of the 1,008 authors here whose files we hold, read 2026-08-06

  • write conventional commit messagesin 36 of 745, across 35 files
  • delete branches after mergein 30 of 745, across 21 files
  • make atomic commitsin 25 of 745, across 15 files
  • write minimal code to pass testsin 22 of 745, across 10 files
  • run tests before committingin 21 of 745, across 13 files
  • re-snapshot after navigation or DOM changesin 21 of 745, across 13 files
  • use try-catch for error handlingin 20 of 745, across 6 files
  • write tests before implementationin 20 of 745, across 8 files
  • configure branch protection rulesin 19 of 745, across 5 files
  • explain the why in commit messagesin 19 of 745, across 9 files
  • refactor code while tests remain greenin 19 of 745, across 6 files
  • Interact with elements using refsin 19 of 745, across 11 files

Said here and by no other author read

  • brief with a goal and done-condition
  • point the agent at relevant files
  • decompose large tasks into reviewed steps
  • make it explore before it edits
  • require the agent to run tests
  • require the agent to run the linter

Grouped from the skills themselves: near-identical wordings counted once, and counted by distinct author, so one author publishing three of these counts once. Length counted with cl100k_base; the agent that loads this file may tokenize it differently.

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.