| name | execplan |
| description | Create living execution documents for complex multi-step tasks. Use for /execplan, "create a plan". NOT for quick tasks or code reviews — only structured phased planning. |
| model | opus |
| effort | high |
| disable-model-invocation | true |
| allowed-tools | Read, Write, Bash, Glob, Grep |
ExecPlan: Living Execution Documents
EXECUTE this skill now. Follow the workflow steps below using the provided $ARGUMENTS. Do NOT describe, summarize, or explain this skill — run it.
Generate, manage, and complete ExecPlans — self-contained, outcome-focused design documents that guide agents through complex tasks.
Constants
TEMPLATE_PATH: C:/Users/gurusharan.gupta/Agents/Claude Code/templates/execplan.md
PRINCIPLES_DIR: C:/Users/gurusharan.gupta/Agents/Claude Code/principles
MANIFEST_PATH: C:/Users/gurusharan.gupta/Agents/Claude Code/manifest.json
Commands
Parse $ARGUMENTS to determine action:
/execplan <description> — Generate a new ExecPlan for the described task
/execplan list — Show active and recently completed plans across all projects
/execplan complete <plan-file> — Mark a plan as complete, fill Outcomes & Retrospective
/execplan (no args) — Show usage help
Workflow: Generate ExecPlan
Step 1: Determine Project Context
Identify the current project. Read its CLAUDE.md if it exists. Understand:
- Tech stack (from package.json, requirements.txt, etc.)
- Directory structure
- Existing architecture patterns
- Active exec-plans (check
docs/exec-plans/active/)
Step 2: Read Template and Principles
Read the ExecPlan template from TEMPLATE_PATH.
Read PRINCIPLES_DIR/_index.json to know which golden principles apply.
Step 3: Research the Task
Before writing the plan, thoroughly research:
- Which files will be affected? Read them.
- What existing patterns should be followed?
- What dependencies are involved?
- What tests exist that might be affected?
- Are there related exec-plans (active or completed)?
Step 4: Generate the ExecPlan
Fill in the template with:
Purpose / Big Picture: What the user can do after this change that they can't do now. How to see it working. User-visible behavior.
Context and Orientation: Current state described for someone who knows nothing. Full paths to key files. Define all terms. No references to external docs — embed everything needed.
Plan of Work: Prose description of edits. Break into milestones if >1 logical phase. Each milestone must be independently verifiable.
Concrete Steps: Exact commands, working directories, expected outputs.
Validation and Acceptance: Observable behavior, not internal attributes.
- "After running X, observe Y"
- "Test Z fails before, passes after"
Interfaces and Dependencies: Prescriptive. Name libraries, types, function signatures.
Idempotence: Steps can be repeated safely. Include recovery for risky steps.
Initialize living sections:
- Progress: Empty checkboxes for each milestone step
- Surprises & Discoveries: "(none yet)"
- Decision Log: "(none yet)"
- Outcomes & Retrospective: "(to be filled at completion)"
Step 5: Ensure Self-Containment
Critical check: Could a novice agent with ONLY this ExecPlan file execute the work end-to-end? If not, add missing context. Do not reference external docs, prior plans, or assumed knowledge.
Key rules:
- Every term of art must be defined in plain language
- Every file path must be full repository-relative
- Every command must include working directory
- Every expected output must be shown
- No "as described previously" or "see the architecture doc"
Step 6: Save
Create directory if needed: docs/exec-plans/active/
Save to: docs/exec-plans/active/{YYYY-MM-DD}-{slugified-title}.md
Filename max 80 characters, lowercase, hyphens.
Step 7: Report
ExecPlan created: docs/exec-plans/active/{filename}
Purpose: {1-sentence summary}
Milestones: {N}
Estimated scope: {files to modify}
Next: An agent can now execute this plan by reading the file and following the milestones.
Workflow: List Plans
When the user runs /execplan list:
- Read the manifest to get all registered projects
- For each project, check for
docs/exec-plans/active/*.md and docs/exec-plans/completed/*.md
- Display:
Active Plans:
Project Plan Created Milestones
───────────────────────────────────────────────────────────────────────────
powers add-api-authentication 2026-03-26 3/5 done
presentation migrate-to-typescript 2026-03-25 0/4 done
Recently Completed:
powers setup-test-infrastructure 2026-03-24 5/5 done
Workflow: Complete Plan
When the user runs /execplan complete <plan-file>:
- Read the plan file
- Verify all Progress checkboxes are checked
- If not all done, ask: "Plan has incomplete steps. Mark as complete anyway?"
- Fill in Outcomes & Retrospective section:
- What was achieved vs. original purpose
- What remains (if anything)
- Lessons learned
- Time from first to last progress timestamp
- Move the file from
docs/exec-plans/active/ to docs/exec-plans/completed/
- Report the move
Important
- ExecPlans are SELF-CONTAINED. A novice with only the plan can execute it.
- ExecPlans are LIVING DOCUMENTS. Update Progress, Surprises, Decision Log as work proceeds.
- ExecPlans are OUTCOME-FOCUSED. Define observable behavior, not code attributes.
- When implementing an ExecPlan: do not prompt for next steps — proceed to the next milestone. Resolve ambiguities autonomously. Commit frequently.
- One ExecPlan per significant task. Don't over-plan trivial changes.
- Plans must be idempotent: steps can be re-run safely.