Correct Course

Run this skill when something has changed — new information, a failed assumption, a reshaped scope — and the current pipeline artifact can no longer be trusted. /correct-course diagnoses the blast radius, names the earliest skill that needs to re-run, and walks you through cleaning up stale artifacts before forward motion resumes.

This is the named, invocable form of the Pipeline Recovery rules in references/SYSTEM-OVERVIEW.md (bundled alongside this skill). Those rules still apply — this skill just gives them a front door.

One question per turn. Throughout this skill, ask one question at a time and wait for the user's answer before asking the next. Backtracking is a conversation, not a form.

Prefer single-select. Use single-select multiple choice when the user is choosing one direction, one priority, or one next step.

Use multi-select rarely. Reserve it for compatible sets — goals, constraints, non-goals, success criteria — that can all coexist. If prioritization matters, follow up asking which selected item is primary.

Use the platform's question tool when available. In Claude Code, use AskUserQuestion; in Codex, request_user_input; in Gemini, ask_user. Otherwise, present numbered options in chat and wait for the user's reply before proceeding.

Invocation Position

This is a side-route skill that reconnects to the main pipeline at whichever earlier skill the diagnosis points to.

Use /correct-course when any of the following is true:

• /research surfaced version changes or constraints that invalidate the /shape closing summary
• /write-a-prd reveals the problem was misframed during shaping
• /prd-to-issues reveals that the total work materially exceeds the stated appetite
• /execute hits an error whose root cause is that the archived research or the PRD is wrong
• /pre-merge's Boundary Map Contracts dimension surfaces a spec-reality mismatch — the code reflects external reality but the declared spec has rotted (non-existent package, renamed public API, version skew past what the research snapshot pinned)
• /pre-merge surfaces an architectural concern that warrants rework, not advisory findings
• A GitHub issue has gone stale because the work it described has been reshaped or superseded

Do not use it for:

• Forward progress that is still on track (use the normal pipeline)
• Bugs that just need triage (use /qa)
• Refactor opportunities unrelated to stale artifacts (use /request-refactor-plan)
• Fresh shaping when no prior artifact exists (use /shape)

Process

1. Capture the trigger

Ask one question: "What changed that made the current artifact untrustworthy?"

Let the user describe it in their own words. Do not ask follow-up questions yet. You are trying to understand the class of invalidation, not the full history:

• New information — a version check, a benchmark, a stakeholder comment, a failed demo
• Failed assumption — something the /shape summary marked Likely or Uncertain turned out to be wrong
• Reshaped scope — the appetite, no-gos, or must-haves need to change
• Wrong problem framing — the user realized they were solving the visible symptom, not the underlying problem
• External force — a dependency was deprecated, a pricing tier changed, a partner API was removed

Write down the class you think this is, in one sentence. You will use it in Step 2.

2. Diagnose the blast radius

Based on the trigger, determine the earliest skill whose output is now untrustworthy. The rule of thumb:

Explore to confirm. Read the PRD issue, any slice issues, the research archive entry for this feature, and the relevant solutions entries. The research entry is at the absolute path the PRD cites (/research Phase 5c states where it is written); when no PRD exists yet, there is no path to cite, so follow /write-a-prd § 4's "Research document" lookup, which says how to match and cite the entry. Find lessons through the bundled resolver references/sidecar.sh: begin the command with SC="<this skill's directory>/references/sidecar.sh" and read every directory bash "$SC" solutions-dirs prints, which owns that policy (its header states it). Check any path the PRD cites with sidecar.sh require <path> --from "PRD #<n>"; a MISSING: line is itself a stale artifact to list in Step 3. The goal is to be specific: not "shaping is stale" but "the Ably-vs-Pusher decision in the research is wrong because Ably raised its free-tier limits and Pusher dropped theirs."

State the diagnosis to the user in one or two sentences and ask one question: "Does that match what changed?" If no, refine. If yes, proceed.

3. List stale artifacts

Now list every durable artifact that needs to be dealt with before forward progress can resume. Be specific — name the exact issue number, the exact file path. Common stale artifacts:

• Research archive entry (the absolute path the PRD cites; /research Phase 5c states where it is written) — if the earliest affected skill is /shape, leave the archive entry in place; it is point-in-time history and cannot mislead downstream skills that no longer reference it. If the earliest affected skill is /research, run /research again to produce a new dated entry superseding the old one; the prior snapshot remains in the archive as context. Archive entries are not deleted during backtracking — they live outside the repo and cost nothing to keep.
• PRD GitHub issue — update in place if the problem and appetite still hold; comment and close if the entire PRD is being reshaped
• Slice GitHub issues for a PRD edited after decomposition — gh issue edit each slice's acceptance criteria and boundary map to match the PRD's current text, then re-run /prd-to-issues § 8 to post a new Decomposed into: comment on the PRD. The new comment is what clears /execute's amendment gate (it compares the PRD's last body edit with the latest such comment); editing the slice alone leaves the gate firing on every later run
• Slice GitHub issues — comment on each and close as superseded if the PRD is being reshaped; comment and leave open only if the specific slice is still valid
• Planning or container milestone — usually leave open; the milestone's feature issues or slice issues are the real artifacts
• Solutions entries (in-tree or in the sidecar) — flag as stale if this situation reveals the documented lesson is wrong, but do not delete without the user's confirmation
• Open PR — close with a comment explaining the backtrack if the PR is built on the stale artifact; leave open if the code is still correct but the framing is wrong
• Active feature branch — usually preserve; the branch is just code history, not a durable planning artifact

For each artifact, name it, name what should happen to it (delete, update in place, close with comment), and note the one-line reason. Present the full list to the user as a numbered plan.

4. Walk the cleanup

Walk the cleanup one artifact at a time. For each:

1. State what you are about to do and why.
2. Ask the user to confirm (one question: "Proceed with this one?").
3. Do the action. Use gh issue close, gh issue comment, or direct file deletion as appropriate.
4. Move to the next.

Do not batch the cleanup into a single confirmation. Each artifact is a small, reversible decision, and batching increases the chance that a subtle mistake slips through. The cost of confirming each is small; the cost of closing the wrong issue is real.

If the user hesitates on any artifact, skip it and flag it as "needs review" in the final summary. Do not press.

5. Hand off

After the cleanup is complete, state explicitly:

• Which skill the user should run next (the earliest affected skill from Step 2)
• What context to bring into that skill (the trigger, the diagnosis, any artifacts preserved for reference)
• Whether the re-run should be scoped to a specific piece (most backtracks are scoped — "re-research the real-time provider decision" — not full re-runs)

Suggest starting the next skill in a fresh session if the conversation has been long. The closing summary is the compressed handoff.

What This Skill is NOT

• Not a bug triage tool. If the trigger is a failing test or a runtime error, use /qa first to diagnose whether it is a bug or a stale-artifact problem (/qa's per-issue depth check delegates to /triage-issue when root cause is needed).
• Not a refactor planner. If the trigger is architectural pain with no stale upstream artifact, use /improve-codebase-architecture or /request-refactor-plan.
• Not a full re-run. A good correction scopes the re-run to what changed, not to everything. "We shaped X, but Y was wrong, so we are re-researching Z" is the target.
• Not a replacement for the Pipeline Recovery reference. The Pipeline Recovery section of references/SYSTEM-OVERVIEW.md is still the canonical description of what backtracking means in Skill Kit — this skill makes that reference invocable. For deeper reading, the repo original is at https://github.com/chrislacey89/skills/blob/main/SYSTEM-OVERVIEW.md.

Handoff

• Expected input: a trigger that invalidated an upstream artifact, plus enough context to diagnose the blast radius
• Produces: a cleaned-up repo state with stale artifacts deleted, updated, or closed, and a named re-entry point into the main pipeline
• Re-enters the main pipeline at: /shape, /research, /write-a-prd, /prd-to-issues, or /execute depending on the diagnosis
• What comes next: the user runs the recommended earlier skill, scoped to the change, in a fresh session if the correction was large