toolkit/packages/skills/workflow/SKILL.md
Guide through structured delivery workflow with plan, implement, validate phases
npx skillsauth add stevengonsalvez/agents-in-a-box workflowInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
You are tasked with orchestrating the complete research->plan->implement workflow for a given task or requirement. This command manages the entire pipeline from initial research through implementation.
When this command is invoked, respond with:
I'll orchestrate the complete workflow for your task.
This will include:
1. **Research**: Comprehensive investigation of the problem space
2. **Planning**: Creating a detailed implementation plan
3. **Implementation**: Executing the plan with verification
Please provide:
- The task/feature/bug you want to address
- Any specific constraints or requirements
- Whether you want to run the full pipeline or start from a specific phase
You can also say "workflow continue" if you have existing research or plans.
Then wait for the user's input.
Before starting any workflow phase, check for persistent project state:
Check for .planning/ directory:
.planning/STATE.md to understand current position.planning/ROADMAP.md for phase contextIf .planning/ does NOT exist:
mkdir -p .planning/phases .planning/todos
{{HOME_TOOL_DIR}}/templates/planning/ to .planning/If .planning/ exists and has ROADMAP.md:
/discuss before planningIf no existing research found in research/ directory:
Invoke research phase:
/research command internallyresearch/YYYY-MM-DD_HH-MM-SS_topic.mdReview research output:
Present research summary:
## Research Complete
Key findings:
- [Major discovery 1]
- [Major discovery 2]
- [Important constraint]
The research has been saved to: research/[filename].md
Shall I proceed to create an implementation plan based on these findings?
After research is complete or if existing research is found:
Invoke planning phase:
/plan command with research contextInteractive planning:
plans/descriptive_name.mdPresent plan summary:
## Plan Complete
Implementation will be done in [N] phases:
1. [Phase 1 name] - [objective]
2. [Phase 2 name] - [objective]
...
The plan has been saved to: plans/[filename].md
Ready to begin implementation?
After plan is approved:
Invoke implementation:
/implement command with plan pathProgress tracking:
Present completion status:
## Implementation Status
Phase 1: [Name] - Complete
Phase 2: [Name] - Complete
Phase 3: [Name] - Partial (see notes)
Verification Results:
- Tests: [status]
- Linting: [status]
- Build: [status]
Would you like me to run validation on the implementation?
If requested or if issues found:
Run validation:
Present validation results:
## Validation Report
Automated Checks:
All tests passing
Linting clean
Build successful
Plan Criteria:
[Criterion 1] - Verified
[Criterion 2] - Verified
[Criterion 3] - Needs manual verification
The implementation is ready for review.
After successful implementation, offer to capture learnings:
## Capture Learnings?
This task is complete. If you encountered any problems worth documenting
for future reference, consider running:
/compound
This will:
- Capture the solution as a searchable learning
- Save it to docs/solutions/ for future sessions
- Make it discoverable via /research
Useful when:
- You debugged a tricky issue
- You found a non-obvious solution
- You want to remember "the fix" for next time
Skip this if the work was straightforward with no notable discoveries.
After each major phase transition, update .planning/STATE.md:
The user can control the workflow with these commands:
workflow start - Begin from research phaseworkflow continue - Resume from existing research/planworkflow skip-research - Start directly with planningworkflow validate - Run validation on completed workworkflow status - Show current progressThroughout the workflow, maintain a master todo list:
Master Workflow: [Task Name]
[ ] Research Phase
[ ] Codebase analysis
[ ] Documentation review
[ ] External research (if needed)
[ ] Synthesize findings
[ ] Planning Phase
[ ] Review research
[ ] Draft plan structure
[ ] Detail each phase
[ ] Define success criteria
[ ] Implementation Phase
[ ] Phase 1: [Name]
[ ] Phase 2: [Name]
[ ] Phase 3: [Name]
[ ] Validation Phase
[ ] Run automated tests
[ ] Verify success criteria
[ ] Generate report
Update this list as you progress through each phase.
If the workflow is interrupted:
Save state:
On resume:
research/plans/User: /workflow
Assistant: I'll orchestrate the complete workflow for your task...
User: Add OAuth2 authentication to the application
Assistant: I'll start by researching OAuth2 implementations and your current auth setup...
[Runs research phase]
[Creates plan based on research]
[Implements with user approval]
[Validates the implementation]
This workflow command internally calls:
/research - For investigation phase/plan - For planning phase/implement - For execution phase/validate - For verification phase (if created)Each command can also be run independently if needed.
documentation
Report reflect drain spend over a time window — tokens split by cached (cache_read), uncached writes (cache_creation), and io (input+output), with a $ estimate, grouped by day / outcome / model / transcript. Reads the drainer's cost log and surfaces outlier runs and cache-reuse health (the 41.5M-token failure mode = low cache reuse + high cache writes). Use to answer "what is reflection costing me" for the last day / week.
development
Show fleet status — every claude session running on the host, merged across ainb + claude-peers broker + background jobs. Use when you need to enumerate sessions before composing an action, see which sessions have a peer registered (broker-routable) vs tmux-only, check the `summary` of each session, or pipe the list into jq for filtering. Default output: text table. Pass --format json for LLM consumption.
testing
Ordered multi-step prompts to fleet targets, ack-gated between steps via JSONL assistant-turn-end detection. Use for cycles like disconnect→reconnect→verify, or any flow where step N+1 requires step N to have completed first. The skill BLOCKS until each target's transcript shows the next assistant turn finishing OR per-step timeout fires (default 300s).
development
Center control panel — enumerate every claude session that is blocked waiting on something: a user answer (AskUserQuestion fired), an API error retry, an idle assistant turn-end with no follow-up, or an explicit WAITING: marker. Returns rich JSON with signal kind + context per session. Use this when you've stepped away from the fleet and want one place to see everything that wants your attention and answer it.