# Audit Before You Build

> Use as the FIRST step of any implementation task handed to you from a checklist, ticket, milestone, or spec — the first step is an existence check, not a design. Trackers record conversations, not deliverables, so grep for the item's key nouns and cross-check spec status before writing code, then pivot "build X" to "verify X and close the gaps" when it already exists. Triggers on "implement this ticket", "build feature X", "next milestone item", "from the checklist", "start this task", "add X".

- Skill: `iamakbarsha1/audit-before-you-build` (Agent Skill)
- Install (CLI): `npx skillmds@latest add iamakbarsha1/audit-before-you-build`
- Raw SKILL.md: https://api.skillmd.com/api/skills/iamakbarsha1/audit-before-you-build/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Security
- Author: iamakbarsha1 (https://skillmd.com/u/iamakbarsha1)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/iamakbarsha1/audit-before-you-build

---


# Audit Before You Build

**Type:** Open-source — client-agnostic methodology, no project-specific detail.

**Created by akbarsha — https://github.com/iamakbarsha1**

Distilled from cases where a checklist item was already ~90% shipped as a side
effect, and where "all issues closed" hid whole unbuilt layers — building from
the tracker text would have duplicated working code or missed real gaps.

**Licence:** Released under CC BY 4.0 — share and adapt for any purpose with
credit. Full text: `LICENSE` at the repository root.

**Feedback & Support:** If a rule here proves wrong or needs sharpening,
open an issue on the repository or contact the author at the profile link
above. If the problem is the agent not following a rule below rather than
the rule itself, that's an execution failure — acknowledge and correct it.

## The core rule

A ticket, checklist, or spec line records a conversation about intent — it is
not a reliable record of what has or hasn't been built. Code drifts from the
tracker in both directions: features ship as side effects of other work
without their item being ticked, and closed items leave real layers unbuilt.
So the first step of "build X" is never design — it's an existence check. Grep
the codebase for the item's key nouns, cross-check the spec's own status
markers, and only then decide whether the task is "build", "finish", or
"verify and close".

## Checks

- **Grep for the item's key nouns before designing.** Take the concrete terms
  from the task — paths, function names, table/column names, endpoint routes,
  spec keywords — and grep for each. Existing hits mean the work is partly or
  wholly done. *(A milestone item was ~90% already implemented as a side
  effect of an earlier change; grepping the feature's route name surfaced it
  immediately, turning a "build" into a small gap-closing task.)*
- **Cross-check the tracker against code and spec status, both ways.** "Closed"
  does not mean "built" and "open" does not mean "absent". Read the spec's own
  Status headers and the code stubs together. *(Illustrative: An "all issues closed"
  milestone hid two entire layers that were never implemented — the tracker
  said done, the code said empty.)*
- **Pivot the verb when reality disagrees.** If the audit shows the thing
  exists, restate the task honestly — "verify the existing X and close gaps A
  and B" instead of "build X" — and say so before writing code. Duplicating
  working code is worse than a slow start. *(Illustrative: A ticket read "build the CSV
  export endpoint"; the audit found a working export handler behind a feature
  flag, so the task was restated as "flip the flag and add the missing
  filename sanitization" before any code was written.)*

## Pre-flight check — before you write implementation code from a task

- [ ] You grepped the codebase for the task's key nouns (paths, names, spec
      terms) and know what already exists.
- [ ] You cross-checked tracker status against actual code AND spec status
      markers, not just the ticket text.
- [ ] If the thing partly exists, you restated the task as verify/finish and
      said so before building.

If any box is unchecked, you're building on the tracker's word — go audit the
code first.