Best fit
Saves a coding agent's working state before it hits the context limit, then resumes in a fresh session without re-explaining everything. Captures the plan, git branch, uncommitted changes, decisions, blockers, and next steps.
// agent productivity · v2.0
HandoffPilot
Create a reviewable handoff before a context reset, resume from the newest saved state with a drift check, or reconstruct a best-effort handoff from transcript cues after an interrupted session.
What it does
A long agent session ends the same way too often: the context window fills, the thread resets, and the next session starts cold. Run /handoff to create a structured Markdown scaffold in .claude/handoffs/ from current repository state; the agent then fills the observed plan, decisions, failed attempts, blockers, open questions, and next steps.
In a fresh session, /resume reads only the newest handoff, summarizes it, asks “Continue where we left off?”, and checks the current branch and worktree for drift after confirmation. It does not commit, push, or alter your code.
New in v2.0: Recovery Mode
Sometimes a session ends without a saved handoff: a crash, a rate-limit stop, or an unexpected reset. Run /recover with a transcript and the current repository to build a best-effort reconstruction. Recovery uses transcript cues, current git state, provenance labels, and candidate-path checks; it is heuristic, so uncertain details still need review.
The original /handoff and /resume work exactly as before. Recovery Mode is a free upgrade for existing customers.
What the handoff can capture
- The active plan and where you are in it.
- The git branch and current uncommitted changes.
- Decisions made so far, so they are not relitigated.
- Open blockers and unresolved questions.
- The concrete next steps.
- A resume prompt the next session can act on immediately.
What the handoff looks like
> /handoff
branch feat/auth-refresh
uncommitted 3 files
blockers token expiry edge case
wrote .claude/handoffs/2026-07-27-handoff.md
next review the scaffold, then use /resume
// choose with context
Is HandoffPilot the right skill?
It returns
Saves a coding agent's working state before it hits the context limit, then resumes in a fresh session without re-explaining everything. Captures the plan, git branch, uncommitted changes, decisions, blockers,….
Do not use it as
Do not treat a successful run as proof beyond the supplied files, declared scope, and evidence the skill actually checked.
Compatibility, access, version, and licence
- Reads or accesses
- Only the files or context you deliberately provide; review the listing before granting additional access.
- Compatibility
- Agents that read the open SKILL.md format. Verify runtime and service requirements at the purchase or install destination.
- Version
- Verify the current package version at the linked destination.
- Licence
- Review the package licence and seller terms at the linked destination.
Questions
What does it capture?
The helper scaffolds current repository state, including git context when available. The agent then fills observed plan progress, decisions, failed attempts, blockers, open questions, and next steps.
Does it commit or change my code?
No. It writes a handoff file describing the state. It does not commit, push, or alter your working tree.
Which agents does it work with?
The workflow targets local SKILL.md-compatible coding agents with file and terminal access, including Claude Code and Codex. The optional helper requires Python 3.10 or newer; Git improves repository-state capture.
Give the next session a state it can inspect.
Current price and purchase options are shown on Agensi. Files stay local; no cloud account or network access is required.
Get HandoffPilot on Agensi