agentleFS
Sign inStart free
GhostwheeIGhostwheeI/Master-Planner/.cursor/skills/project-memory/SKILL.md

project-memory

Skill0 starsChanged 5 months agoMITView on GitHub

Maintains durable, inspectable planning memory and enables session resume. Use to record decisions/facts/bugs/pitfalls so they survive compaction, to check prior decisions before proposing architecture, and to recover an in-progress plan after context loss. Operates on .cursor/plans/<project>/ files.

24 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

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

  • 0

    Common

    Most of the 19 edited versions we read keep these.

  • 5

    Distinctive

    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:

    Keep it terse and human-legible.

    line 20
  2. Very specific:

    Before proposing architecture, read Key Decisions in context.md.

    line 24
  3. Distinctive:

    Before re-fixing a bug or re-attempting a failed approach, read Known Pitfalls.

    line 25
  4. Very specific:

    Before continuing an in-progress plan, read manifest.md.

    line 26
  5. Very specific:

    Update manifest.md whenever the plan changes phase (phase, status, next action, active assumptions, resume instruction).

    line 30
  6. Distinctive:

    Update memory only when it earns its place.

    line 32
  7. Distinctive:

    Never: write secrets (keys, tokens, passwords, PII) into any file.

    line 33
  8. Distinctive:

    If you suspect context was lost (compaction, new session, very long conversation):

    line 37
  9. Very specific:

    Read .cursor/plans/<project>/manifest.md first — it contains the current phase, active task, open assumptions, and an explicit resume …

    line 38
  10. Very specific:

    Read spec.md and tasks.md.

    line 39

“Common” is measured: kept by at least 6 in 10 of the 19 edited versions of this file we could read. “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

Project Memory & Resume

Chat memory is volatile; files are durable. This skill keeps the small amount of state that must survive compaction on disk, and defines how to recover from it.

Adapts the flat-markdown "memory lanes" + read-before-act protocol from SpillwaveSolutions/agent_rulez project-memory (license unconfirmed — pattern adapted, not copied; see docs/ATTRIBUTION.md) and the session-checkpoint resume pattern from russelleNVy/three-man-team (MIT).

The memory bank lives in the plan directory

State for a project is under .cursor/plans/<project>/:

  • manifest.md — the checkpoint: phase, status, active assumptions, open questions, resume instruction, decision log.
  • context.md — the memory bank: Glossary, Key Facts (no secrets), Key Decisions (ADR-style), Known Pitfalls, Retrospective Notes.
  • spec.md, tasks.md — the canonical plan.

Keep it terse and human-legible. The goal is a grep-friendly safety net, not a second codebase.

Read-before-act protocol

  • Before proposing architecture, read Key Decisions in context.md. Do not re-open a settled decision recorded as an ADR without flagging that you are doing so and why.
  • Before re-fixing a bug or re-attempting a failed approach, read Known Pitfalls.
  • Before continuing an in-progress plan, read manifest.md.

Write-at-boundaries protocol

  • Update manifest.md whenever the plan changes phase (phase, status, next action, active assumptions, resume instruction).
  • Append to context.md when a decision/fact/pitfall emerges that will prevent a future mistake. Use the ADR format in ../master-planner/references/ADR-FORMAT.md for significant decisions; use the glossary format in ../master-planner/references/CONTEXT-FORMAT.md for terminology.
  • Update memory only when it earns its place. Avoid churn.
  • Never write secrets (keys, tokens, passwords, PII) into any file.

Resume after context loss

If you suspect context was lost (compaction, new session, very long conversation):

  1. Read .cursor/plans/<project>/manifest.md first — it contains the current phase, active task, open assumptions, and an explicit resume instruction.
  2. Read spec.md and tasks.md.
  3. Continue from the recorded phase. Do not restart and do not invent state.

The /resume command and the optional preCompact hook (see .cursor/hooks/) both point here.

Anti-Rationalization

ExcuseReality
"I'll keep it in my head"Compaction erases your head. The file survives.
"Recording decisions is bureaucracy"One ADR line stops a future you from re-deciding it wrong.
"I'll update the manifest at the end"If context is lost before the end, there's nothing to resume from. Update at boundaries.
"More memory files = better memory"Churn buries signal. Keep four lanes, terse.

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.

Check your setup, free