agentsclimarketplace

Pwrl end session checkpoint

Skill wicttor/pwrl/pwrl-end-session-checkpoint

Plan. Work. Review. Learn. — A minimal disciplined agentic development framework.

Install
npx -y skills add wicttor/pwrl --skill pwrl-end-session-checkpoint

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

2 things to look at

  • no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
  • 4 stars4 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Verify repository state and confirm session completion before committing.

SKILL.md

6.0 KB, ~1.3k tokens by cl100k_base, as published. Nobody here has run it

PWRL End Session Checkpoint

Verify repository state and get explicit user confirmation that the session is ready to end.

Interaction Method

  • Use platform's ask_user_question, ask_user, ask_user_input, vscode/askQuestions or any available extension/tool for user interaction for all decisions
  • Ask one question at a time
  • Use multiple-choice questions when possible
  • If input is empty, ask: "Is the repository state ready to be checkpointed? Review any pending changes before confirmation."
  • Provide clear recovery suggestions when errors occur

Purpose

Checkpoint is the first phase of session cleanup. It verifies that the repository has changes to commit, reviews what will be included, and confirms with the user that work is complete and ready to finalize:

  • Check working tree state (modified, staged, untracked files)
  • Report changes that will be included in the commit
  • Get user confirmation that session is ready to close
  • Handle incomplete work (ask for next steps to document)
  • Generate checkpoint artifact for Phase 2 commit creation

Usage

/pwrl-end-session-checkpoint                              # Default: check current state
/pwrl-end-session-checkpoint "switching to feature-X"     # With context

This is typically invoked by /pwrl-end-session orchestrator, or directly to verify state before committing.

Support Files

  • references/checkpoint-protocol.md — Detailed validation rules and decision tree

Detailed Workflow

For complete step-by-step instructions, see checkpoint-protocol.md.

This SKILL.md provides an overview. The detailed workflow document contains:

  • Working tree detection and file categorization
  • Change summary and confirmation flow
  • Session reason capture and next steps
  • Version bump detection
  • Checkpoint artifact generation
  • Error handling and recovery

Quality Gate Validation

After completing this phase, run quality gate validation:

/pwrl-phase-checkpoint end-session 1 [artifact-path]

See pwrl-phase-checkpoint for validation rules.


Workflow

Phase 1: Verify Working Tree

Detect all changes and determine if commit is viable.

See detailed workflow: checkpoint-protocol.md

  • Run git status to detect modified, staged, untracked files
  • If no changes found, skip the mode ask (Step 1.5) and exit — there is nothing to commit
  • Categorize changes into staged, unstaged, untracked

Step 1.5: Select Interaction Mode

If changes were detected, ask the user to choose their engagement level for this session-end. Use the platform's ask_user_question extension (or equivalent) to present the following three options:

Question: "How would you like to proceed with this session-end?"

Options:

  • Detailed (Step-by-Step) — Show the draft commit message in pwrl-end-session-commit and let the user edit it before approval; show pre-flight summary in the checkpoint. Maximum control. Best for complex sessions with breaking changes, version bumps, or sensitive attribution.
  • Smart (Risk-gated automation) — Show a pre-flight summary (files, line counts, version-bump check) and approve the commit with one click; only pause for HIGH-risk operations (e.g., version bump detected, breaking-change warning). v1 simplification: behaves like Yolo with a single confirmation prompt at workflow start.
  • Yolo (Full Automation) — Entire session-end (checkpoint + commit) auto-runs and only reports the final commit SHA. Fastest. Best for routine, low-risk session ends.

Store the selection in the checkpoint artifact (replacing the placeholder in the schema above):

interactionMode: detailed | smart | yolo

Empty working tree: If there are no changes, the mode ask is skipped (since there is nothing to commit). Users with no changes will not be prompted.

Learnings chaining: The optional Phase 3 chain to /pwrl-learnings has its own mode ask (set by pwrl-learnings-extract Step 1.5). The mode set in this step controls only checkpoint and commit behavior.

Phase 2: Review and Confirm Changes

Get user explicit approval on what will be committed.

See detailed workflow: checkpoint-protocol.md

  • Display summary of all changes
  • Ask user to confirm this list is correct
  • Optionally allow inclusion/exclusion of specific files
  • Ask user to describe why work is ending here
  • If incomplete: ask for next steps to document

Phase 3: Generate Checkpoint Artifact

Create structured output for Phase 2 (commit creation).

See detailed workflow: checkpoint-protocol.md

  • Structure approved files as array
  • Detect version bump (check package.json, VERSION files)
  • Create session summary with reason and next steps
  • Validate all required fields
  • Generate YAML artifact with metadata

Rules

  • ✓ Exit if no changes (no point committing empty session)
  • ✓ Always show user what will be committed before proceeding
  • ✓ Always get explicit user confirmation (never auto-proceed)
  • ✓ Handle incomplete work gracefully (ask for next steps)
  • ✓ Detect version bumps (will inform commit message in Phase 2)

Best Practices

  • Be explicit about what files are included (show paths clearly)
  • If work is incomplete, make next steps actionable and specific
  • Indicate if this is cleanup of abandoned work vs. paused progress
  • Note if any breaking changes or significant decisions were made

Acceptance Criteria

  • ✓ Input: Repository has changes to commit
  • ✓ Output: Checkpoint artifact with approved_files, session_reason, user_confirmation
  • ✓ Verification: User explicitly confirmed files and reason
  • ✓ Version detection: Bumped version (if any) identified for Phase 2

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.