Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add Jamie-BitFlight/claude_skills --skill rebasegit clone --depth 1 https://github.com/Jamie-BitFlight/claude_skillsWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/jamie-bitflight/claude_skills/rebase)<a href="https://agentmods.dev/skills/jamie-bitflight/claude_skills/rebase"><img src="https://agentmods.dev/badge/skills/jamie-bitflight/claude_skills/rebase/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/jamie-bitflight/claude_skills/rebase"><img src="https://agentmods.dev/badge/skills/jamie-bitflight/claude_skills/rebase.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00102 | $0.03095 |
| Opus 5.5 | $0.00041 | $0.01238 |
| Sonnet 5.5 | $0.00020 | $0.00619 |
| Haiku 4.5 | $0.00010 | $0.00310 |
Grade A, and why
rebase scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 20d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
Copies of this mod
1 near-identical copy found in the catalogue:
- rebase — 100% identical, 205 lines differ
How it starts
The opening of the file, as written. The whole thing — 140 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Keywords: rebase, git rebase, history replay, rebase conflict, continue rebase, abort rebase, git worktree, rewritten history, authorized force-with-lease publication
Result-bound verification
For start/continue completion, bind validation to the immutable result commit R.
Run repository-required and changed-interaction checks on a clean detached worktree
or equivalent isolated checkout of R, separately from restored user changes.
Preserve unrelated work; a passing dirty worktree does not validate its commit.
Record the checked OID and environment; verify checks left tracked source unchanged.
Source-mutating checks invalidate that evidence. Commit a required correction within
the existing task authority, then rebind R and rerun its checks. When that commit
is blocked or unauthorized, pause rather than reporting the correction published.
Every result change, including remote reconciliation, invalidates earlier result
validation. Reobserve the named result ref before completion; it must still name
R. A decision or stopped-state path cannot become a success merely because its
refs or ancestry still satisfy the goal.
Publication authority
A start request to rebase a branch authorizes publishing R to that branch's own remote
destination, using --force-with-lease bound to an observed destination OID. Local-only
wording in the request withdraws it. Continue publishes only under authority bound at
start; abort never publishes. Any other destination, any lease not bound to an
observed OID, and any other force-push needs explicit authority.
Other parties update remote feature branches. Fetch the destination before replay and again before pushing; integrate what it gained, then continue (publication).
Workflow
flowchart TD
Start([Agent receives explicit start, continue, or abort request]) --> Kind{Request?}
Kind -->|Start| Bind[Agent: from request/task context bind exact refs/OIDs, requested observable history/ref relation, destinations, authority, worker facts]; Kind -->|Continue or abort| Active["`Agent: read [active recovery](./references/active-rebase-recovery.md); bind metadata-owning worktree/Git dir`"]
Bind --> BindResult{All dependent facts observed or explicitly bound?}
BindResult -->|Yes| Satisfied{No active rebase; S/T unchanged; ancestry-only; T ancestor of S?}; BindResult -->|Missing or ambiguous| Decision; BindResult -->|Failure or unobservable| Recover
Satisfied -->|Yes; destinations equal required result; read-only| NoChange[Agent: record no-change evidence]; Satisfied -->|No or transformation goal| Locate
Satisfied -->|Yes; only publication differs| Locate; Satisfied -->|Failure or unobservable| Recover
Active --> ActiveResult{Inspection result?}; ActiveResult -->|No metadata in bound worktree and Git dir| NoActive[Agent: record absent metadata and no mutation]; ActiveResult -->|Owner and required contract rebound| Safe
ActiveResult -->|Required contract fact missing or inconsistent| Decision; ActiveResult -->|Failure or unobservable| Recover
Locate[Agent: find exact source attachment; record worktree/Git-dir ownership] --> Location{Bound source location?}
Location -->|Source branch attached here| Safe; Location -->|Source branch unattached| BranchWT[Agent: create start-only branch worktree]
Location -->|Source attached elsewhere| BindWT[Agent: bind that exact worktree/Git dir]; Location -->|Non-branch OID; result bound| DetachedWT[Agent: create start-only detached worktree]
Location -->|Failure or unobservable| Recover; BindWT --> Safe
BranchWT --> WTResult{Creation result?}; DetachedWT --> WTResult
WTResult -->|HEAD OID equals source; branch attachment matches when required| Safe; WTResult -->|Failure or unobservable| Recover
Safe[Agent: observe worktree writers; request pause when supported] --> Owner{Worker checkpoint observation?}
Owner -->|No writer, acknowledged pause, or worker ended| Operation{Operation?}; Owner -->|Writer active| Wait[Agent: wait once for pause, end, or command result]
Owner -->|Unknown or unreachable| Decision
Wait --> WaitResult{Observed wait outcome?}; WaitResult -->|Checkpoint met or worker ended/failed without writer| Operation; WaitResult -->|Still running and cannot pause/end| Decision
WaitResult -->|Failure or unobservable| Recover
Operation -->|Start| Prepare[Agent: observe status; checkpoint only task-authorized changes after worker checkpoint]; Operation -->|Continue or abort| ActiveGuard[Agent: reobserve same owner metadata and saved-entry identity]
ActiveGuard --> ActiveGuardResult{Owning worktree/Git-dir metadata unchanged?}
ActiveGuardResult -->|Continue| Stop{Current active-rebase state?}; ActiveGuardResult -->|Abort| Abort["`Agent: read [active recovery](./references/active-rebase-recovery.md); abort in bound owner and restore pre-state`"]
ActiveGuardResult -->|No, failure, or unobservable| Recover
Prepare --> PrepResult{Status empty after checkpoint containing every observed change?}
PrepResult -->|Yes| Refs; PrepResult -->|No; checkpoint commit blocked| Save["`Agent: read [named stash](./references/named-stash.md); save and bind exact entry`"]
PrepResult -->|Failure or unobservable| Recover
Save --> SaveResult{Exact entry verified?}; SaveResult -->|Yes| Refs; SaveResult -->|No or unobservable| Recover
Refs["`Agent: reobserve exact source and target; when a publication destination is bound, fetch it per [publication](./references/publication.md) remote sync`"] --> RefResult{Ref result?}
RefResult -->|Bound OIDs unchanged| Next; RefResult -->|Same named refs or bound destination moved| Account[Agent: account for movement; integrate compatible destination commits into the source before replay]
RefResult -->|Failure or unobservable| Recover
Account --> AccountResult{Every moved commit has one evidence-backed disposition?}
AccountResult -->|All classified; no supported conflict| Next; AccountResult -->|Unclassified or conflicting| Decision
AccountResult -->|Failure or unobservable| Recover
Next{Reobserved start path?} -->|Replay required| Shape; Next -->|No-replay publication; bound goal relation true| Current[Agent: bind current result OID R]
Next -->|Failure or unobservable| Recover
Current --> CurrentResult{Named result ref resolves to R?}; CurrentResult -->|Yes| Saved; CurrentResult -->|No or unobservable| Recover
Shape{Merge in fresh replay set?} -->|Yes| History["`Agent: read [history shape](./references/history-shape.md); choose state-specific disposition`"]
Shape -->|No| StartReplay[Agent: start fresh replay in bound start worktree]; Shape -->|Failure or unobservable| Recover
History --> HistoryResult{Supported state-specific disposition?}
HistoryResult -->|Fresh start topology chosen| StartReplay; HistoryResult -->|Active commit: skip| Skip[Agent: skip only the classified active commit]
HistoryResult -->|Active commit: preserve| Preserve[Agent: preserve classified active commit, then continue]; HistoryResult -->|No or conflicting| Decision
HistoryResult -->|Failure or unobservable| Recover
Stop -->|Unmerged entries| ReplayConflict; Stop -->|Topology, equivalent, or empty| History
Stop -->|Metadata active; resolution staged or no conflict| ContinueReplay[Agent: continue the bound active replay]
Stop -->|Other failure or unobservable| Recover
StartReplay --> ReplayResult{Replay operation result?}; ContinueReplay --> ReplayResult; Skip --> ReplayResult; Preserve --> ReplayResult
ReplayResult -->|Metadata/unmerged absent; R bound; result ref and bound goal relation match| Saved{Lifecycle saved entry?}
ReplayResult -->|Unmerged entries| ReplayConflict["`Agent: read [conflict and ambiguity](./references/conflict-and-ambiguity.md); resolve replay intent`"]
ReplayResult -->|Topology, equivalent, or empty| History; ReplayResult -->|Other failure or unobservable| Recover
ReplayConflict --> ReplayIntent{Exactly one outcome preserves compatible intent, passes checks, and is staged?}
ReplayIntent -->|Yes; active metadata unchanged| ContinueReplay; ReplayIntent -->|None or incompatible alternatives| Decision
ReplayIntent -->|Failure or unobservable| Recover
Abort --> AbortResult{Abort restoration predicate?}
AbortResult -->|Metadata/unmerged absent; source or HEAD equals pre-replay OID| Saved; AbortResult -->|Failure or unobservable| Recover
Saved -->|Verified none| FinishMode{Lifecycle outcome?}; Saved -->|Exact entry bound| Restore["`Agent: read [named stash](./references/named-stash.md); apply exact OID and preserve entry`"]
Saved -->|Ambiguous or unobservable| Recover
Restore --> RestoreResult{Apply result?}
RestoreResult -->|Conflict-free| RestoreFinish[Agent: verify tree/checks; relist unique entry; drop selector; confirm absent]
RestoreResult -->|Conflict; entry preserved| RestoreConflict["`Agent: read [conflict and ambiguity](./references/conflict-and-ambiguity.md); resolve restoration intent`"]
RestoreResult -->|Other failure or unobservable| Recover
RestoreConflict --> RestoreIntent{Exactly one outcome preserves compatible intent and passes checks?}
RestoreIntent -->|Yes| RestoreFinish; RestoreIntent -->|None or incompatible alternatives| Decision
RestoreIntent -->|Failure or unobservable| Recover
RestoreFinish --> RestoreFinishResult{Checks zero, no unmerged entries, exact entry absent?}; RestoreFinishResult -->|Yes| FinishMode; RestoreFinishResult -->|No or unobservable| Recover
FinishMode -->|Explicit abort| OrientAbort[Agent: inventory interactions; verify restored assumptions]
FinishMode -->|Start or continue| Orient[Agent: bind committed result R; run repository and changed-interaction checks on its isolated tree]
Orient --> VerifyResult{Checks pass on unchanged tracked tree of R; result ref equals R; no unmerged entries; intersections recorded?}
VerifyResult -->|Yes| Publish{Publication bound and authorized?}
VerifyResult -->|Failure attributable to replay; one correction within bound goal| Correct[Agent: apply and commit only the task-authorized replay correction]; VerifyResult -->|Not attributable, outside goal, missing intent, or alternatives| Decision
VerifyResult -->|Failure or unobservable| Recover
Correct --> CorrectResult{Correction committed within task authority?}
CorrectResult -->|Yes; invalidate old evidence and rebind R| Orient; CorrectResult -->|Commit blocked or unauthorized| Decision; CorrectResult -->|Failure or unobservable| Recover
Publish -->|No| Handoff; Publish -->|Yes| Remote["`Agent: read [publication](./references/publication.md); reconcile destination movement for this attempt`"]
Remote --> RemoteResult{Remote stage result?}
RemoteResult -->|Final fetch unchanged; exact lease current; R revalidated; result ref equals R| Push[Agent: push immutable R with the authorized exact lease]
RemoteResult -->|Destination moved at final observation| Retry
RemoteResult -->|Unexpected result-ref change| Decision
RemoteResult -->|Conflict| RemoteConflict["`Agent: read [conflict and ambiguity](./references/conflict-and-ambiguity.md); resolve remote intent`"]
RemoteResult -->|Failure or unobservable| Recover
RemoteConflict --> RemoteIntent{Exactly one outcome preserves compatible intent and passes checks?}
RemoteIntent -->|Exactly one| Retry; RemoteIntent -->|None or incompatible alternatives| Decision
RemoteIntent -->|Failure or unobservable| Recover
Push --> PushResult{Push result?}
PushResult -->|Exit zero; post-fetch destination equals R| Handoff; PushResult -->|Lease rejected or destination moved| Retry{Attempts used below the cap in publication.md?}
Retry -->|Yes| Remote; Retry -->|No| LeaseStop[Agent: stop external mutation; report each attempt, its rejection or movement, and the destination]
PushResult -->|Other failure or unobservable| Recover
NoChange --> Handoff; NoActive --> Handoff; Decision[Agent: stop mutation; report missing fact/alternatives and evidence] --> Handoff
Recover["`Agent: read [active recovery](./references/active-rebase-recovery.md); stop mutation and collect report`"] --> Handoff
OrientAbort --> Handoff; LeaseStop --> Handoff
Handoff[Agent: carry the reached lifecycle path; check acquired worker obligation] --> Worker{Worker obligation?}
Worker -->|None| Terminal{Reached lifecycle path and current evidence?}; Worker -->|Acquired| Deliver[Agent: deliver summary and resume/stop decision]
Worker -->|Unknown or unobservable| Pending([Agent: stop mutation; preserve repo; report last worker state, handoff attempt, missing ack/observation; no completion])
Deliver --> HandoffResult{Acknowledged summary/resume, or worker observed stopped with handoff?}
HandoffResult -->|Yes| Terminal; HandoffResult -->|No or unobservable| Pending
Terminal -->|Abort request and restored pre-state| Aborted([Aborted and restored])
Terminal -->|Successful start/continue path; validation current for R; result ref and post-push destination equal R| Published([Published completion])
Terminal -->|Successful start/continue path; validation current for R; result ref equals R; goal relation true; no publication| Local([Local completion])
Terminal -->|No-change evidence complete| NoChangeDone([No change]); Terminal -->|Metadata absent; no mutation| NoActiveDone([No active rebase])
Terminal -->|Decision report complete| Paused([Paused for decision]); Terminal -->|Stopped-state report complete| Stopped([Stopped with observed state])
Terminal -->|Inconsistent or unobservable| Recover
What ships with it
6 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 20d ago Changed · +12 lines · +14 tokens per session da2465bc3ff3
- 28d ago Changed · +17 tokens per session 58963ce1b7fe
- 29d ago Changed · +100 lines · -2 tokens per session 3b59c4cb02db
- 1mo ago Changed · -63 lines · -41 tokens per session a90bfd00aeb3
- 1mo ago First seen · 91 lines · 114 tokens per session scan A 6f68c7633650
rebase is a skill published in the GitHub repository Jamie-BitFlight/claude_skills (67 stars, last pushed 4d ago), licensed MIT. It adds 102 tokens to every session and 3,095 once invoked, about $0.0004 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
merge-batch
Prepare and merge an owner-selected PR batch with contributor credit and exact-candidate admission. Use for a pasted PR list; combined focused validation is diagnostic, and every actual landing still requires its applicable checks and authorization.
prowler-commit
Creates professional git commits following conventional-commits format. Trigger: When creating commits, after completing code changes, when user asks to commit.
gh-auth-isolation
Safely manage multiple GitHub identities (EMU + personal) in agent workflows.
github-skill
Work with GitHub via the gh CLI — clone repositories, create/list/merge pull requests, create/list issues, and run any other gh command (API calls, workflow runs, releases, repo administration). List operations return parsed JSON.
comet-github
A routing guide for Comet-related GitHub work. It directs requests about pull requests, issues, CI failures, ideas, and fixes to the appropriate review or implementation process.
release-helper
Ensures configuration and code changes are released correctly. Use PROACTIVELY whenever you edit config.json, change any setting, or fix a configuration bug in this project, so the change reaches production.