Hyper plan review
Use after a plan exists in .hyperclaude/plans/ and should be sanity-checked before execution. Also when the user invokes /hyperclaude:hyper-plan-review. Runs Codex critique on the plan — distinct from /hyperclaude:hyper-code-review (code, not plans).From its SKILL.md
npx -y skills add zeikar/hyperclaude --skill hyper-plan-reviewAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 5 stars5 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.
SKILL.md
5.2 KB, as published. Nobody here has run it
hyper-plan-review
Plan review gate. Locates a plan file, sends it to Codex for critique, saves the review to .hyperclaude/plan-reviews/<timestamp>-<slug>.md, and you read the review and refine the plan.
When to use
- User typed
/hyperclaude:hyper-plan-review(with or without an argument). - You just produced an implementation plan and want a critic pass before executing.
How to invoke
Invocation argument: $ARGUMENTS
--resume is supported. Paths with spaces are unsupported.
Argv grammar
Apply this regex to the trimmed $ARGUMENTS:
^(?:((?!--)\S+))?(?:\s*(--resume)(?:\s+(\S+))?)?\s*$
- Group 1 = optional plan path (negative lookahead prevents matching
--resumeas a path) - Group 2 = literal
"--resume"token (truthy when present, undefined when not) - Group 3 = optional resume artifact path
When Group 2 is '--resume' (truthy) and Group 3 is undefined, treat as --resume auto.
Valid invocations:
/hyperclaude:hyper-plan-review— auto-discovers most recent plan, fresh run/hyperclaude:hyper-plan-review path/to/plan.md— explicit plan path, fresh run/hyperclaude:hyper-plan-review --resume— auto-discovers plan, resumes from latest artifact (auto)/hyperclaude:hyper-plan-review --resume <prev-artifact-path>— resumes from explicit artifact/hyperclaude:hyper-plan-review path/to/plan.md --resume— explicit path, resume fromauto/hyperclaude:hyper-plan-review path/to/plan.md --resume <prev-artifact-path>— explicit path, explicit artifact
If the argument doesn't match the regex, ask the user to clarify and stop.
Resume semantics
See ${CLAUDE_PLUGIN_ROOT}/references/bridge-review-calls.md for the shared --resume semantics (explicit vs auto fallback, the template-version precondition). Plan-review's identity check keys on the plan path.
- Budget exceeded (docs > 200KB after revision): bridge returns
ok:false— NOT fallback. Tell user to narrow scope.
Step 1 — Resolve the plan path
In priority order:
-
Explicit argument (Group 1). If Group 1 is non-empty, treat it as the plan path.
-
Most recent in
.hyperclaude/plans/. If Group 1 is empty, run via the Bash tool:ls -1t .hyperclaude/plans/*.md 2>/dev/null | head -1If that returns a path, use it.
-
Nothing found. Tell the user: "No plan file found. Write your plan to
.hyperclaude/plans/<slug>.md, or pass an explicit path:/hyperclaude:hyper-plan-review path/to/plan.md." Stop.
Step 2 — Compose the review brief (or omit it)
Compose per ${CLAUDE_PLUGIN_ROOT}/references/review-brief.md, assigning the scratchpad path to BRIEF_FILE per its shell-safety recipe; with no admissible source, omit the flag.
Concrete omission case for this skill: a fresh session with no conversation history that auto-discovers a plan (Step 1.2) has no (a)/(b) source at all — omit the flag, or ask the user for their requirements and use their reply. Never mine the plan file's prose — it is Claude-authored; quoting it would let the planner bless its own additions.
Step 3 — Run the bridge
node "${CLAUDE_PLUGIN_ROOT}/scripts/codex-bridge.mjs" plan-review --plan-path "<resolved path>" [--review-brief "$(cat "$BRIEF_FILE")"]
If --resume was matched (Group 2 truthy), append --resume <value> to the argv passed to the bridge, where <value> is Group 3 if present, otherwise auto. The --review-brief "$(cat "$BRIEF_FILE")" token is OPTIONAL — present when a source exists (fresh and --resume alike), omitted otherwise.
Use the Bash tool with timeout: 600000 and run_in_background: true (invocation mode: ${CLAUDE_PLUGIN_ROOT}/references/bridge-review-calls.md).
Step 4 — Surface the review
Once the bridge task completes, parse the JSON envelope from its stdout per ${CLAUDE_PLUGIN_ROOT}/references/bridge-review-calls.md. On success, read the review file with the Read tool. Integrate the review:
- Blockers / Majors — fix the plan before proceeding.
- Minors / Improvements — optionally fold in.
- Verdict — communicate to the user.
When you revise the plan, always overwrite the same plan file in place. Do not create -v2.md or sibling files: the review thread's --resume identity is keyed on the plan path, so a new path breaks resume continuity (the next critique loses its prior context). Tell the user the plan was updated in place.
Output contract
Review files have frontmatter (mode: plan-review, task: plan path, slug, generated, plugin-version, codex-version, template-version, plan-path, review-brief (present when a brief was supplied, or carried forward from a successfully-resolved resume), plus codex-input-tokens, codex-cached-input-tokens, codex-output-tokens, codex-reasoning-output-tokens each emitted independently when Codex reported that token field in usage; omitted when Codex did not emit usage) followed by Issues / Improvements / Verdict sections.
What ships with it
Read from the repository
Just SKILL.md. No reference files, no scripts.
Gives 0 of the 12 instructions most plan spec skills give
Counted across 1,360 of the 2,617 authors here whose files we hold, read 2026-09-06
- Ask one question at a timein 73 of 1360
- Write the spec using the templatein 22 of 1360
- Ask clarifying questions if neededin 19 of 1360, across 18 files
- Wait for user confirmation before proceedingin 19 of 1360
- Save plans to the plans directoryin 17 of 1360, across 13 files
- Check for product marketing context firstin 16 of 1360, across 5 files
- Read the plan file completelyin 16 of 1360
- Order tasks by dependencyin 16 of 1360
- Gather context from the conversationin 15 of 1360, across 9 files
- Explore the codebase instead of askingin 15 of 1360, across 13 files
- Wait for explicit user approvalin 14 of 1360, across 13 files
- Quiz the user on the breakdownin 13 of 1360, across 7 files
Said here and by no other author read
- Send the plan to Codex for critique
Grouped from the skills themselves: near-identical wordings counted once, and counted by distinct author, so one author publishing three of these counts once.