agentleFS
yiyaw-labyiyaw-lab/agent-armor/continuity-context/SKILL.md

continuity-context

Skill1 starsChanged 5 months agoMITView on GitHub

Maintains CONTEXT.md and session handoff rules so agents and humans resume work without tool sprawl. Invoke explicitly when the user mentions lost context, context switching, where were we, handoff, or working across IDE/PM/notes tabs — not on generic "continuation" phrasing alone.

The file

We split this skill into the rules it gives an agent

7 rules. About 0 are common, 5 are distinctive, and 2 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.

  • 2

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Keep ≤40 lines. Session-end update takes <30 seconds.

    line 20
  2. Very specific:

    If CONTEXT.md contradicts git/branch state, trust git and fix CONTEXT in the same edit.

    line 42
  3. Very specific:

    If CONTEXT.md missing and project has >1 day of work → create from git log + open files.

    line 57
  4. Distinctive:

    Not a replacement for tickets/docs — it's the pointer layer

    line 67
  5. Distinctive:

    No automated append on every keystroke — batch at meaningful checkpoints only

    line 68
  6. Distinctive:

    No MCP catalog, Slack sync, or multi-tool dashboards

    line 69
  7. Distinctive:

    If Continuum project open: checkpoint mental notes may live in graph — CONTEXT.md is for cross-IDE portability only

    line 70

≈ “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

Continuity Context

Goal

One file + one habit so the next session (human or agent) starts in <30 seconds — without syncing Jira, Notion, and the IDE separately.

Artifacts

CONTEXT.md (repo root)

Keep ≤40 lines. Session-end update takes <30 seconds.

# Context

**Active goal:** One line — what "done" looks like this week.

**Active work:** WI/ticket name or branch (link if exists).

**Open decisions (max 3):**
- [ ] Decision — options A/B

**Touched today:** (paths, newest first, max 8)
- src/...

**Next action:** One imperative verb — "Ship X-3 IPC tests" / "Fix Track drag".

**Last session note:** One sentence — where you stopped.

Rotate Touched today — drop paths older than 3 days unless still hot.

If CONTEXT.md contradicts git/branch state, trust git and fix CONTEXT in the same edit.

User rule snippet (optional)

Offer this for Cursor Rules (≤10 lines):

At task start: read CONTEXT.md if present; restate Active goal + Next action in one line.
At meaningful checkpoints (task done, PR opened, context switch): batch-append changed paths to Touched today (max 8).
Before ending a session: update Last session note + Next action only.
Do not ask the user to maintain a separate PRD unless they request it.

Workflow

  1. If CONTEXT.md missing and project has >1 day of work → create from git log + open files.
  2. At session start → read it; don't re-scan whole repo unless user asks.
  3. At session end → update Last session note + Next action only (minimal diff).
  4. At checkpoints (task complete, PR ready, deliberate pause) → batch-update Touched today; do not append on every edit or keystroke.
  5. Do not duplicate full architecture in CONTEXT — link the blueprint layer:
    • Continuum: .continuum/blueprint/ or BlueprintPanel (not ARCHITECTURE.md)
    • Other repos: ARCHITECTURE.md if the architecture-blueprint skill exists

Constraints

  • Not a replacement for tickets/docs — it's the pointer layer
  • No automated append on every keystroke — batch at meaningful checkpoints only
  • No MCP catalog, Slack sync, or multi-tool dashboards
  • If Continuum project open: checkpoint mental notes may live in graph — CONTEXT.md is for cross-IDE portability only

Adversarial self-check

  • New teammate could orient from CONTEXT.md alone in 30s?
  • File still ≤40 lines?
  • Next action is one verb, not a paragraph?
  • Session-end edit touched only Last session note + Next action (not a full rewrite)?
  • Avoided duplicating blueprint/architecture content?

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.