Review every changed file in its own fresh context

• .claude/skills/review-each-file/SKILL.mdorchestrator
• .claude/skills/review-each-file/graph.skill.mdgraph source
• .claude/agents/review-each-file-scope.mdsubagent
• .claude/agents/review-each-file-review-file.mdsubagent
• .claude/agents/review-each-file-merge.mdsubagent
• .claude/agents/review-each-file-gate.mdsubagent
• .skill-graph/.gitignoresupport
• .cursor/rules/review-each-file.mdcCursor rule
• README.mdreadme

---
name: review-each-file
description: "Lists the files a change touches, reviews each one in its own fresh-context run, merges the findings and gates them before reporting."
---

# review-each-file

Lists the files a change touches, reviews each one in its own fresh-context run, merges the findings and gates them before reporting.

This skill is a graph: each step has its own prompt and context, and only the payloads named below move between steps.

## Plan

Run the stages in order. Every model step runs in its own subagent; do not do a step's work yourself. Only command steps and approvals are yours to run. Steps in the same stage do not depend on each other: start them together, in one message with several subagent calls.

The input (`goal`) is the user's goal: their request, word for word, handed on under `### goal`. Ask for it if the user has not given one.

### Stage 1 · 1 step

- **LIST THE FILES** (`scope`) → subagent `review-each-file-scope`
  - Hand it: the user's goal.
  - Expect: `file_list` (json).
  - Contract check (`file_list`): save its output to `.skill-graph/review-each-file/scope.json`, then run `node -e "const fs=require('fs');const d=JSON.parse(fs.readFileSync(process.argv[1],'utf8'));const k=['files','focus'];const ok=d!==null&&typeof d==='object'&&Array.isArray(d)===false;const m=ok?k.filter((x)=>Object.hasOwn(d,x)===false):k;if(m.length>0){console.error('Missing keys: '+m.join(', '));process.exit(1)}" .skill-graph/review-each-file/scope.json` with Bash. Exit code 0 passes.
  - If a contract check fails: re-run this step once with its previous answer and the error output. If it fails again, stop and show the user the output.
  - Hand-off by file: save its output, verbatim, to `.skill-graph/review-each-file/scope.json`. Later steps get that path, not the text.

### Stage 2 · 1 step

- **REVIEW EACH FILE** (`review_file`) → subagent `review-each-file-review-file` · fresh context · one run per item, up to 6 at a time
  - Take the list: the `files` array in the JSON of LIST THE FILES (`scope`). If it is not an array, stop and report. If it is empty, skip this step: its output is `[]`.
  - If the list has more than 40 items, stop before starting any run and tell the user how many there are.
  - Start one run of the subagent per item, up to 6 at a time (several subagent calls in one message); start the next as one finishes.
  - Hand each run: its item: one entry of `files` from the JSON of LIST THE FILES (`scope`); the `focus` value from the JSON of LIST THE FILES (`scope`), as the path `.skill-graph/review-each-file/scope.json`. Nothing else: no notes of your own, no other step's reasoning, no earlier conversation.
  - Expect, per run: the result for its item. Checks and gates below apply to each run on its own; a retry re-runs only the runs that failed.
  - Output: a JSON array of the runs' results, in item order. That array is what later steps get.
  - Hand-off by file: save its output, verbatim, to `.skill-graph/review-each-file/review_file.json`. Later steps get that path, not the text.

### Stage 3 · 1 step

- **MERGE FINDINGS** (`merge`) → subagent `review-each-file-merge`

… 37 more lines in the kit