AGENT SKILL

chat-closeout

Assess or resolve a chat's unfinished work and readiness to leave on read, start a new topic, or archive. Use for conversation-state questions and requested closeout, including selected-project or global scope.

License 许可
MIT
最近更新
2026-11-14
Stars
★ 1
Forks
0

01 前言 Front Matter

name
chat-closeout
description
Assess or resolve a chat's unfinished work and readiness to leave on read, start a new topic, or archive. Use for conversation-state questions and requested closeout, including selected-project or global scope.

02 资源文件 1

03 SKILL.md 正文

Chat Closeout

Let the user leave the conversation without abandoning work or losing necessary material. One skill handles assessment and execution; closeout includes its own assessment and verification.

Follow the request

“What's outstanding?” or “Can I archive this?” asks for a read-only assessment. “Close this out” or a bare invocation asks to resolve unfinished business within existing authority. “Close this out and archive it” also requests the archive after readiness is established. Preparation alone does not request archival. Do not require a sequence of skill invocations.

Default to this chat. For a broader scope, enumerate the requested accessible chats with pagination and disclose omissions; global scope does not expand project authority. Inspect archived chats only when requested or relevant to a dependency. Investigate deletion only when explicitly requested, verifying independent preservation and the host's deletion effects separately.

Establish the completion boundary

Read original requests and corrections back to the beginning or a verified checkpoint, carrying its outstanding obligations forward. An earlier “done” claim is not such a checkpoint by itself. Reconcile relevant project records, artifacts, actual results, and receiving owners; disclose consequential missing history. Reuse adequate evidence without repeating an entire audit.

Determine what finishing the user's intended outcome requires, separately from whether each action is authorized. Repository closeout normally includes committing accepted changes and completing their intended integration and synchronization, alongside required checks and installation. Do not classify these outcomes as optional just because the user did not repeat each Git verb. An explicitly chosen local-only result, deliberate deferral, or effective transfer can establish a different endpoint; unrelated dirty files do not block this task.

Required work awaiting permission remains unfinished. Carry existing authorization forward; do not ask again when it already covers the action. A closeout request permits routine preservation and reconciliation within scope, but does not waive review holds or override publication, messaging, deletion, or spending restrictions. Finish independent authorized work and prepare the concrete remaining action before asking for genuinely missing authority. Never obtain a YES by silently dropping, deferring, or declaring a requirement unnecessary.

Find consequential unfinished requests, deliverables, verification, integration/publication, promises, human decisions, delegated supervision, and owned resources. Distinguish direct human instructions from quotations, proposals, agent promises, and automated submissions. Silence, a read receipt, and a finished model turn establish neither approval nor completion. Optional ideas need no disposition unless they resolve an existing obligation.

Resolve or preserve unfinished business

During assessment, leave files, tasks, chats, schedules, and running work unchanged. During closeout, choose the useful disposition for each obligation:

  • Finish and verify authorized work, attributing changes and preserving unrelated work. Do not expand cleanup into new features or whole-repository housekeeping.
  • Transfer only when authorized, with effective receiving ownership, context, authority, next action, and responsibility for supervision, integration, and acceptance. Delivery of a message alone does not establish transfer.
  • Preserve required decisions or deliberate deferrals in the existing authoritative tracker, with consequences and an effective discovery/follow-up path. A buried note, unnamed future reader, or promised report back only here is insufficient. Do not invent a schedule or turn optional suggestions into commitments.
  • Preserve necessary decisions, evidence, authorization sources, and artifacts in their appropriate durable homes. The repository owns continuity for repository work; another chat or harness must be able to resume from its records without needing this transcript. An uncommitted file or installed copy alone is not evidence of backup or completed integration.
  • Retire owned temporary resources within authority, retaining or transferring those needed by continuing work. Do not stop or delete resources merely because they were mentioned.

For every consequential open item, establish its owner, status, next action or trigger, holds, and discoverable destination. Work remains here unless finished or supported by a verified continuation independent of the user returning to this exchange. A recorded trigger and an actual monitor are different things.

Verify and report

Check evidence for the outcome being claimed. Matching installed files proves installation consistency, not commitment or synchronization. For required Git or external operations, verify the intended changes, revisions, and destinations, including relevant uncommitted or untracked work. A record declaring “complete” is a conclusion to substantiate, not independent proof. Confirm no obligation lost its owner, changed status without authority, or disappeared from scope.

Report two visible verdicts using YES, NO, or UNVERIFIED. Known unmet conditions mean NO; missing consequential evidence, including an unresolved completion boundary, means UNVERIFIED.

Action Conditions for YES
Ready to leave on read / start a new topic Agreed work is verified complete or has effective continuation elsewhere. No obligation depends on the user remembering to return here; required human input has a discoverable destination and appropriate follow-up.
Ready to archive The leave condition holds, and no work, follow-up, or resource requires this chat to remain active or visible. Necessary historical references remain retrievable under verified host archive behavior.

Separately say whether all agreed work is finished or work continues elsewhere. Name outstanding human and agent actions, owners, direct evidence/destination links, and what would change a non-YES verdict. Ask a concrete question when a required decision remains; completed work and optional ideas need no acknowledgment. State actual commit/sync/archive results separately from readiness, with scope, inspection time, and material limits. Do not hide unfinished work behind a YES or imply that unrelated repository work is published.

When archival is requested, perform it through supported controls after readiness is established and report the observed result. Preserve the completion record outside the target chat if the action would otherwise remove the only usable record. A failed action remains outstanding. Do not mark chats read or archive them merely to obtain a clear report.

Readiness is a snapshot; new work or failed follow-up can invalidate it. Standing and task chats share the criteria, regardless of whether the user intends to keep them open. Reuse does not erase context. Prefer existing records and tools; do not create a new global task system or append optional work as a closure condition.

04 同仓库更多 Skills

chrispalmo
chrispalmo/agent-config

search-agent-chats

Search retained Cursor, Codex, and Claude Code conversations for past work, decisions, and context when historical recall would help. Exposes archive scope, freshness, coverage, and event citations. Do not run at session start.

★ 1
chrispalmo
chrispalmo/agent-config

plaintext

Produce one copyable plain-text payload only when the human explicitly invokes plaintext. Use for prompts, messages, emails, and other requested copy-paste output.

★ 1
chrispalmo
chrispalmo/agent-config

interview

Run an adaptive context-gathering interview when the user explicitly invokes interview before a decision or task. Reuse known answers and stop when the decision-critical context is sufficient.

★ 1
chrispalmo
chrispalmo/agent-config

github-repo-defaults

Apply GitHub ownership and Snowflake1 connection defaults when creating a new repository or publishing a local project to a new GitHub repository. Use for repository creation requests, including project setup that creates a repo.

★ 1
chrispalmo
chrispalmo/agent-config

github-init

Initializes the current project as a Git repository, creates a GitHub repository, and pushes the initial branch with upstream tracking. Defaults to a private repository in snowflake1xyz with Snowflake1 guidance connected; explicit text may override owner or visibility. Use only when the human explicitly invokes github-init.

★ 1
chrispalmo
chrispalmo/agent-config

decision-criteria

Assess a decision or revisit a prior recommendation against the user's objectives, criteria, constraints, and evidence. Use only when explicitly invoked.

★ 1