Daily Handoff

Daily Handoff

Create one compact multi-task recovery checkpoint. Do not implement, verify, review, close, commit, or change task authority.

Canonical usage:

$daily-handoff GAME-201 GAME-202 GAME-203 GAME-204
$daily-handoff

1. Resolve tasks exactly

1. Use the current repository unless --repo or --project identifies one

exact path or unique repository name. Never guess between repositories.

1. Read repository AGENTS.md, CLAUDE.md, their declared order, task/session

conventions, and existing production/session-state/active.md.

1. Resolve each complete ID only as production/tasks//contract.md and an

explicit in-repository contract.md path directly. Preserve input order and deduplicate exact IDs. Never fall back to similarly named stories or partial IDs.

1. With no task arguments, use the same active-task selection as task-status:

lifecycle state in_progress, reopened, blocked, or ready_to_close plus tasks named by active orchestration. Do not include untouched ready backlog or closed tasks.

1. A missing or ambiguous task receives a BLOCKED handoff section with the

exact identity problem; it is not reinterpreted as prose.

2. Collect attributable evidence read-only

Read each contract's managed lifecycle block, implementation/verification/ bugfix handoffs, ownership/worktree ledgers, associated branch/base revision and diff, retained verification artifacts, the complete managed defect ledger, and Git/Plastic metadata. For a legacy task without the block, read sibling status.md only as compatibility input; never create, update, migrate, or delete it.

For every task report exactly:

• Status: recorded lifecycle status and current phase, labeling inference;
• Completed work: only outcomes proven by a handoff, diff, or closure record;

for defect work, include only rows with fresh VERIFIED evidence;

• Changed systems: contract area plus attributable paths/assemblies/assets;
• Verification evidence: check, result, source, revision, and freshness;
• Defects: unresolved and verified counts, stable IDs, linked-record integrity,

and stale proof;

• Remaining blocker: one concrete blocker or None;
• Next command: one short ID-first command or focused unblock action;
• Commit status: uncommitted, committed , none, or UNKNOWN.

Do not attribute root dirty work to a task without worktree/branch/handoff proof. Mark unsupported claims UNKNOWN and outdated evidence STALE. Do not run tests/builds, refresh/import Unity, use Unity MCP/UnitySkills to manufacture evidence, mutate Editor state, start or interrupt agents, create commits, push, close tasks, or change external trackers.

Use short navigation:

• recorded ready_to_close: $task-done ;
• new unrecorded post-implementation bug feedback: $task-bug [] "";
• any populated defect-ledger row, or a lifecycle/handoff that records a

reopened or resumable cycle: $task-cycle ;

• an empty defect ledger with initial ready work or incomplete initial

implementation/evidence: $implement-task ;

• active normal work: Wait for ;
• blocked: one exact decision or unblock action.

Do not route a defect-free initial implementation handoff to task-cycle merely because implementation evidence exists. Once a defect row exists or the task has entered a reopened/resumed cycle, keep its continuation on task-cycle unless ready_to_close takes precedence. When a workflow is already actively writing the task, use Wait for rather than a duplicate invocation.

3. Update only the managed session checkpoint

Write only the managed daily-handoff block in production/session-state/active.md:

# Daily Handoff — 

- Repository: 
- Snapshot revision: 
- Snapshot fingerprint: 
- Updated at: 

## 

- Status: 
- Completed work: 
- Changed systems: 
- Verification evidence: 
- Defects:  |  | 
- Defect records: 
- Remaining blocker: 
- Next command: ``
- Commit status: 

Preserve the status-line block, manual notes, session extracts, and all content outside these markers. Create the file/directories when absent. Keep exactly one managed block and update only requested task sections while retaining unselected task sections from the same active checkpoint.

Before writing, compare the rendered task state and snapshot fingerprint with the existing managed block. If unchanged, preserve the file byte-for-byte, including timestamp. If changed, replace the managed block atomically and re-read it. The checkpoint is recovery state, not authority: never edit a contract or its lifecycle block, legacy status.md, production/tests/assets, ownership, VCS, or trackers.

4. Report

Return the same concise task blocks in chat and finish with:

DAILY HANDOFF
Tasks: 
Checkpoint: production/session-state/active.md | unchanged | not written
Snapshot: 
Decisions required: 
Next: 

Do not start a task-cycle or claim completion from missing/stale evidence. Project documentation is English unless repository rules say otherwise.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: furkantokkan
• Source: furkantokkan/agent-foundry
• License: MIT

Install and usage instructions live in the source repository linked above.