Skills-compatible agents / Implementation workflow
Executing Plans in the Current Session
Runs an approved plan task by task in the current agent session, carrying decisions and verification evidence in a persistent ledger.
315,272 repository stars · checked 2026-11-10. Count applies to obra/superpowers, not this individual skill.
Guide by IDAKIM Digital. Source review: . Original explanation based on the linked publisher sources.
What it does
The reviewed workflow is the inline implementation option in Superpowers. One agent executes tasks rather than paying for a fresh implementer and reviewer for every task; a fresh whole-branch review remains at the end when supported. The plan and its specification remain the authority. A workspace ledger stores completion ranges, test results and explicit rulings so an interrupted or compacted session can resume without rebuilding completed work. Companion helper scripts prepare task briefs and record passing checks; reading this catalogue does not run them.
When to use it
- Implement an approved multi-task feature using one continuous session.
- Resume a long implementation after context loss without repeating completed tasks.
- Track a necessary deviation when repository reality contradicts a plan step.
Step-by-step workflow
- Read the specification and plan, verify workspace isolation and inspect an existing ledger before starting new work.
- Check interfaces shared between tasks, recording any conflict and the specification-based ruling.
- Read the exact task brief, observe its failing tests and implement the smallest behavior satisfying the specified checks.
- Run the complete task verification and record evidence before marking completion; carry every deviation in the ledger.
- Perform the final whole-branch review, resolve material findings with regression evidence and hand off integration choices.
Example request
Use executing-plans for the approved billing-export plan. Work in the existing authorized isolated checkout. Resume only tasks without a verified completion entry. Read each task brief, run its red/green checks and log test output. If a signature conflicts with the approved specification, record the ruling and its risk. Keep customer data out of diagnostics. Finish with the required final review; leave publishing to the already agreed release process.
Worked example
Task 3 might consume export rows defined by Task 1. If Task 1 produced an immutable tuple but the brief assumes a mutable list, implementation should consult the specification and record the chosen interface in the ledger. Quietly adapting one task leaves the next task exposed to the same mismatch. Completion needs the named test command and observed outcome. On resumption, the recorded commit range and tests explain what is done more reliably than an agent’s remembered summary.
What you get
Implemented tasks with tests, a durable completion-and-rulings ledger and a final review record tied to the actual change range.
How to evaluate the result
- Completed ledger entries match real changes and fresh test output.
- Every deviation from the plan has a reason, consequence and carried decision.
- The final review covers the whole branch, including cases individual task tests missed.
What you need
An approved plan and reachable specification, Git history, runnable tests and a supporting coding harness. The full upstream workflow references bundled helper scripts and a reviewer facility.
Setup and activation
The publisher documents separate installation routes for Claude Code, Codex App and CLI, Cursor and other harnesses. Use the README for your exact client; installing the whole Superpowers framework can activate additional workflows automatically. Preserve this skill’s companion files and relative references. Review and security-scan the exact selected version before installation. This directory only reviews sources; it has not installed, executed or tested this skill on those clients.
Limits and considerations
Inline execution saves repeated contexts but gives up independent per-task review. The source requires one final reviewer when available and discloses self-review when unavailable. Its helpers and cleanup behavior need inspection before execution. A passing task is not authorization for a merge, publish or destructive operation; follow actual user permissions and preserve the ledger until final handoff.
Related workflows
- Use writing-plans to prepare unambiguous task briefs first.
- Use verification-before-completion to assess every completion claim.
License and reuse
MIT, copyright 2025 Jesse Vincent. Preserve the copyright and permission notice when redistributing substantial source portions; inspect separately supplied dependencies and assets under their own terms.
Further reading
Read the original source
Publisher: obra/superpowers
Source reviewed: 2026-11-10
Source reviewed; not installed, executed or security-scanned for this directory. These example requests were written by IDAKIM Digital. Upstream content can change.