rekey-work-items
Skill34 starsChanged 51 days agoMITView on GitHub
Converge every work item's ID onto the configured work.id_pattern,
In this repository38 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- hooks/CLAUDE.mdCLAUDE.md
- tasks/CLAUDE.mdCLAUDE.md
- clean-commentsSkill
- raise-prSkill
- configureSkill
- initSkill
- migrateSkill
- pathsSkill
- create-adrSkill
- extract-adrsSkill
- review-adrSkill
- analyse-design-gapsSkill
- inventory-designSkill
- describe-prSkill
- respond-to-prSkill
- review-prSkill
- create-noteSkill
- create-planSkill
- implement-planSkill
- review-planSkill
- stress-test-planSkill
- validate-planSkill
- conduct-spikeSkill
- research-codebaseSkill
- research-issueSkill
- research-topicSkill
- commitSkill
- visualiseSkill
- create-work-itemSkill
- extract-work-itemsSkill
- list-work-itemsSkill
- refine-work-itemSkill
- review-work-itemSkill
- stress-test-work-itemSkill
- sync-work-items-recoverySkill
- sync-work-itemsSkill
- update-work-itemSkill
The file
We split this skill into the rules it gives an agent
6 rules. About 0 are common, 3 are distinctive, and 3 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈3
Distinctive · estimate
This author's own choices. Worth a look.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 19
If no "Agent Names" section appears above, use these defaults: accelerator:reviewer, accelerator:codebase-locator, …
- Distinctive: line 72
Exit 0 with no lines: every item already matches the pattern.
- Very specific: line 104
When stderr carries warning: the working copy has uncommitted changes, name the paths listed under it and say the apply will refuse until …
- Very specific: line 111
When stderr carries
with ACCELERATOR_FORCE set, accelerator work rekey will apply over them, name the paths listed under it and say the … - Distinctive: line 160
Never offer it under the override note.
- Distinctive: line 182
If any invocation exits non-zero, that item violates the canonical frontmatter standard.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Re-key Work Items
!accelerator config context --skill rekey-work-items --fail-safe
!accelerator config agents --fail-safe
If no "Agent Names" section appears above, use these defaults: accelerator:reviewer, accelerator:codebase-locator, accelerator:codebase-analyser, accelerator:codebase-pattern-finder, accelerator:documents-locator, accelerator:documents-analyser, accelerator:web-search-researcher.
Active integration: !accelerator config work integration --fail-safe
Local ID prefix: !accelerator config work key --fail-safe
ID pattern: !accelerator config work id_pattern --fail-safe
Work items directory: !accelerator config path work --fail-safe
/rekey-work-items gives every work item an id in the shape ID pattern
describes, whatever shape it has now. Each re-keyed item is renamed, its old
id joins its aliases, and references to it across every corpus root
follow. The accelerator work rekey engine owns every decision: which items
change, the order fresh numbers follow, collisions, and recovery. This skill
previews, asks, then applies or discards.
A preview records a proposed run plan under .accelerator/state/rekey/,
which version control ignores. The apply re-keys exactly the previewed
mapping, refusing if any item's identity changed since, and records its
progress first, so an interrupted run is finished by running it again. Under
a {tracker} pattern the preview reads the tracker for current keys; the
apply makes no tracker call. The apply refuses over uncommitted changes in
the corpus, because version control is how a re-key is undone.
Pass --per-item-reads through to the preview when the user gives it.
The override note
ACCELERATOR_FORCE makes work rekey skip its safety checks: the apply's
working-copy check, and the discard's check that every path the run wrote is
restored. Never set it, by prefix, export or env: overriding a
refusal is the user's decision, made outside this skill.
Whenever a work rekey call's stderr carries a line starting
warning: ACCELERATOR_FORCE is set, on any exit, tell the user, before any
question, that ACCELERATOR_FORCE is set in their environment and the next
apply or discard will skip its safety check. From then on, for the rest of
this invocation, never offer Undo it: say the override would discard the
run without checking the restore, and ask the user to unset
ACCELERATOR_FORCE and run /rekey-work-items again.
Step 1: Preview
accelerator work rekey --preview [--per-item-reads]Branch on the outcome:
- Exit 0 with mapping lines on stdout, one
<old> → <new>per changing item: go to Step 2. - Exit 0 with no lines: every item already matches the pattern. Report that there is nothing to re-key and stop.
E_REKEY_IN_PROGRESS, after the stderr linethe run in progress has still to re-key:: an earlier run is unfinished. Show the entries it still has to re-key, printed on stdout. Use theAskUserQuestiontool with Finish it (runaccelerator work rekey, then act on Step 3's outcomes) and Undo it (the Undo flow below). Under the override note, offer Finish it and Stop instead.E_REKEY_PLAN_UNREADABLE: relay the printed remedy and any half-applied evidence lines as they are. Never offer to finish the run, and never run the forced discard the remedy names. Stop.- Another exit 4: surface the code, its printed remedy and the paths listed under it. Nothing was recorded. Stop.
- Exit 70: the tracker was unreachable; nothing was recorded. Retry once it is reachable.
- Exit 72-74: the tracker cannot be selected or configured. Relay the
printed message, point at
/accelerator:configure, and stop; nothing was recorded. - Exit 1 with
E_IDENTITY_RUN_BUSY,E_REKEY_STATE_UNWRITABLE,E_REKEY_IDENTITY_RUN_LOCK_FAILED,E_REKEY_CORPUS_INACCESSIBLEorE_DRAFT_ID_EXHAUSTED: relay the printed remedy and stop.
Step 2: Confirm
Print the mapping lines as they are, in a text fence, with the override
note if it applies.
With no dirty-path warning, use the AskUserQuestion tool:
- Yes, re-key — rename the listed items and rewrite their references
- No, discard — discard the plan; nothing changes
When stderr carries warning: the working copy has uncommitted changes,
name the paths listed under it and say the apply will refuse until they are
committed or reverted. Then use the AskUserQuestion tool:
- Keep the plan and stop — commit or revert the paths first
- Discard — discard the plan; nothing changes
When stderr carries with ACCELERATOR_FORCE set, `accelerator work rekey`
will apply over them, name the paths listed under it and say the apply
will rewrite them over the uncommitted changes, which version control then
cannot undo. Then use the AskUserQuestion tool:
- Keep the plan and stop — commit or revert the paths first
- Yes, re-key over them — apply despite the uncommitted changes
Step 3: Apply
On Yes, re-key:
accelerator work rekeyOn exit 0, each stdout line is <old> → <new> and the re-keyed file's
path, separated by a tab. Report each re-keyed item, then go to Step 5. On a
non-zero exit, act on the outcome table; any lines printed before the
refusal are entries that were applied, and they stand.
| Code | What to do |
E_REKEY_PREVIEW_STALE | The work items changed since the preview; name what changed and offer to run /rekey-work-items again for a fresh preview |
E_REKEY_DIRTY_WORKING_COPY | Name the paths, say the plan is kept as it was (proposed or in progress), and stop |
E_REKEY_WORKING_COPY_UNPROBEABLE | Name the failure, say the plan is kept as it was, and stop |
E_REKEY_ENTRY_ROLLED_BACK | Say an entry was rolled back; offer to run accelerator work rekey again to resume |
E_REKEY_ENTRY_REFUSED, E_REKEY_RESTORE_INCOMPLETE | Relay the printed remedy, then offer the Undo flow, unless the override note was shown |
E_REKEY_NOT_REVERTED | The Undo flow, unless the override note was shown |
E_REKEY_PLAN_UNREADABLE, E_IDENTITY_RUN_BUSY | Relay the printed remedy and stop |
| Any other code | Surface the code and its printed remedy, and stop |
Step 4: Discard or keep
On No, discard or Discard:
accelerator work rekey --discardReport that no work item changed.
On Keep the plan and stop: say that once the paths are committed or
reverted, /rekey-work-items previews again and asks, or
accelerator work rekey applies this plan if no item's identity changed
meanwhile.
Undo flow
Never offer it under the override note.
accelerator work rekey --discardOn exit 0 the run plan is gone; report that the re-key was undone. On
E_REKEY_NOT_REVERTED, some paths the run wrote still differ from their
state when it started. Show the paths under each restore from <revision>
heading with that revision, and the paths under remove. The skill does not
restore them: ask the user to restore and remove them, and to say when that
is done. Then run --discard again. If it refuses again, show only the paths
still listed and repeat.
Step 5: Validate
Validate each re-keyed item, using the path each applied line prints:
accelerator corpus frontmatter validate --file <each re-keyed item path>If any invocation exits non-zero, that item violates the canonical frontmatter standard; report the emitted violation before completing.
!accelerator config instructions rekey-work-items --fail-safe
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

