agentleFS
delorenjdelorenj/ssbnk/.agents/skills/plan-protocol/SKILL.md

plan-protocol

Skill2 starsChanged 51 days agoMITView on GitHub

Guidelines for creating and managing implementation plans with citations

The file

We split this skill into the rules it gives an agent

10 rules. About 0 are common, 5 are distinctive, and 5 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈5

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 5

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    When creating or updating a plan, ensure:

    line 12
  2. Distinctive:

    After receiving a complex user request

    line 27
  3. Distinctive:

    When tracking progress across phases

    line 28
  4. Distinctive:

    After research that informs architectural decisions

    line 29
  5. Very specific:

    Use plan_save with this exact markdown format:

    line 42
  6. Very specific:

    Use delegation_list() to see all delegations

    line 145
  7. Very specific:

    Use delegation_read("id") to verify content before citing

    line 146
  8. Distinctive:

    Error: Only one task may be marked CURRENT.

    line 214
  9. Very specific:

    Error: Use [COMPLETE], not [DONE].

    line 233
  10. Very specific:

    Before calling plan_save, verify:

    line 252

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Plan Protocol

Load this skill when creating or updating implementation plans.

TL;DR Checklist

When creating or updating a plan, ensure:

  • YAML frontmatter with status, phase, updated
  • ## Goal section (one sentence)
  • ## Context & Decisions table with citations (ref:delegation-id)
  • Phases with status markers: [COMPLETE], [IN PROGRESS], [PENDING]
  • Tasks with hierarchical numbering (1.1, 1.2, 2.1)
  • Only ONE task marked ← CURRENT
  • Citations for all research-based decisions

When to Use

  1. Starting a multi-step implementation
  2. After receiving a complex user request
  3. When tracking progress across phases
  4. After research that informs architectural decisions

When NOT to Use

  1. Simple one-off tasks → use built-in todos instead
  2. Pure research/exploration → use delegations only
  3. Quick fixes that don't need tracking
  4. Single-file changes with no dependencies

Plan Format

Use plan_save with this exact markdown format:

---
status: STATUS
phase: PHASE_NUMBER
updated: YYYY-MM-DD
---

# Implementation Plan

## Goal
ONE_SENTENCE_DESCRIBING_OUTCOME

## Context & Decisions
| Decision | Rationale | Source |
|----------|-----------|--------|
| CHOICE | WHY | `ref:DELEGATION_ID` |

## Phase 1: NAME [STATUS_MARKER]
- [x] 1.1 Completed task
- [x] 1.2 Another completed task → `ref:DELEGATION_ID`

## Phase 2: NAME [IN PROGRESS]
- [x] 2.1 Completed task
- [ ] **2.2 Current task** ← CURRENT
- [ ] 2.3 Pending task

## Phase 3: NAME [PENDING]
- [ ] 3.1 Future task
- [ ] 3.2 Another future task

## Notes
- YYYY-MM-DD: Observation or decision `ref:DELEGATION_ID`

Frontmatter Fields

FieldValuesDescription
statusnot-started, in-progress, complete, blockedOverall plan status
phaseNumber (1, 2, 3...)Current phase number
updatedYYYY-MM-DDLast update date

Phase Status Markers

MarkerMeaning
[PENDING]Not yet started
[IN PROGRESS]Currently being worked on
[COMPLETE]Finished successfully
[BLOCKED]Waiting on dependencies

State Machine

Plan Lifecycle

not-started → in-progress → complete
                         ↘ blocked

Phase Lifecycle

[PENDING] → [IN PROGRESS] → [COMPLETE]
                         ↘ [BLOCKED]

Task Lifecycle

[ ] unchecked → [x] checked

Critical Rules

  1. Only ONE phase may be [IN PROGRESS] at any time
  2. Only ONE task may have ← CURRENT marker at any time
  3. Move ← CURRENT immediately when starting a new task
  4. Mark tasks [x] immediately after completing them

Citations & Delegations

Where Citations Come From

Citations reference delegation research. The flow is:

  1. You delegate research: delegate to researcher or explore
  2. Delegation completes with a readable ID (e.g., swift-amber-falcon)
  3. You cite that research in the plan: ref:swift-amber-falcon

When to Cite

SituationAction
Architectural decision based on researchAdd to Context & Decisions table
Task informed by researchAppend → ref:id to task line
Implementation detail from researchInline citation in Notes

How to Find Delegation IDs

  • Use delegation_list() to see all delegations
  • Use delegation_read("id") to verify content before citing

❌ NEVER

  • Make up delegation IDs
  • Cite without actually reading the delegation
  • Skip citations for research-based decisions

Examples

✅ CORRECT: Well-formed plan

---
status: in-progress
phase: 2
updated: 2026-01-02
---

# Implementation Plan

## Goal
Add JWT authentication with refresh token support

## Context & Decisions
| Decision | Rationale | Source |
|----------|-----------|--------|
| Use bcrypt (12 rounds) | Industry standard, balance of security/speed | `ref:swift-amber-falcon` |
| JWT with refresh tokens | Stateless auth, mobile-friendly | `ref:calm-jade-owl` |

## Phase 1: Research [COMPLETE]
- [x] 1.1 Research auth patterns → `ref:swift-amber-falcon`
- [x] 1.2 Evaluate token strategies → `ref:calm-jade-owl`

## Phase 2: Implementation [IN PROGRESS]
- [x] 2.1 Set up project structure
- [ ] **2.2 Add password hashing** ← CURRENT
- [ ] 2.3 Implement JWT generation

## Phase 3: Testing [PENDING]
- [ ] 3.1 Write unit tests
- [ ] 3.2 Integration tests

## Notes
- 2026-01-02: Chose bcrypt over argon2 for broader library support `ref:swift-amber-falcon`

❌ WRONG: Missing frontmatter

# Implementation Plan

## Goal
Add authentication

Error: Plan must have YAML frontmatter with status, phase, updated.

❌ WRONG: Multiple CURRENT markers

## Phase 2: Implementation [IN PROGRESS]
- [ ] **2.1 Task one** ← CURRENT
- [ ] **2.2 Task two** ← CURRENT

Error: Only one task may be marked CURRENT.

❌ WRONG: Decision without citation

## Context & Decisions
| Decision | Rationale | Source |
|----------|-----------|--------|
| Use Redis | It's fast | - |

Error: Decisions must cite research with ref:delegation-id.

❌ WRONG: Invalid phase status

## Phase 1: Research [DONE]

Error: Use [COMPLETE], not [DONE]. Valid markers: [PENDING], [IN PROGRESS], [COMPLETE], [BLOCKED].


Troubleshooting

Error MessageFix
"Missing frontmatter"Add ---\nstatus: in-progress\nphase: 1\nupdated: 2026-01-02\n--- at top
"Multiple CURRENT markers"Remove ← CURRENT from all but the active task
"Invalid citation format"Use ref:delegation-id format (e.g., ref:swift-amber-falcon)
"Missing goal"Add ## Goal section with one-sentence description
"Empty phase"Add at least one task to each phase
"Invalid phase status"Use [PENDING], [IN PROGRESS], [COMPLETE], or [BLOCKED]

Before Saving Checklist

Before calling plan_save, verify:

  • Frontmatter: Has status, phase, and updated date?
  • Goal: Is there a clear, one-sentence goal?
  • Citations: Are all research-based decisions cited with ref:id?
  • Single CURRENT: Is exactly one task marked ← CURRENT?
  • Valid markers: Do all phases use valid status markers?
  • Hierarchical IDs: Are tasks numbered correctly (1.1, 1.2, 2.1)?

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.