agentleFS
kevinmanasekevinmanase/continuity-agent-skills/skills/continuity/SKILL.md

continuity

Skill1 starsChanged 6 months agoMITView on GitHub

Use when starting or resuming Continuity-tracked agent work, selecting next work, fetching agent context, finishing/blocking work, resolving review feedback, merging/pushing/rebasing PRs, making significant progress on a tracked node, or before a final response when code, PR state, merge state, blockers, or verification changed.

  • Installs packages

7 other projects use a version of this.

Adopted by

Most recently taken up by

The file

Show its rules and text

We split this skill into the rules it gives an agent

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

  • 0

    Common

    Most of the 5 edited versions we read keep these.

  • 5

    Distinctive

    This author's own choices. Worth a look.

  • 10

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Use the continuity CLI against the hosted API.

    line 10
  2. Very specific:

    If the continuity command is missing, ask the user to install it:

    line 14
  3. Distinctive:

    If the user has a private hosted origin, save it through login:

    line 26
  4. Very specific:

    After login succeeds, rely on ~/.continuity/config.json.

    line 32
  5. Very specific:

    Use the current agent name instead of codex when running in another agent.

    line 45
  6. Very specific:

    Read the returned sequence or continuation before choosing work.

    line 47
  7. Distinctive:

    Follow the packet's authority boundary, stop conditions, verification guidance, and final-answer reconciliation instructions.

    line 53
  8. Very specific:

    Use insert when a user asks for new work, reports a bug, makes a durable decision, or provides an actionable task that should become part …

    line 57
  9. Very specific:

    Use intake only when the input should stay in a pending review queue instead of directly mutating graph work:

    line 63
  10. Distinctive:

    After inserting graph work, rerun:

    line 69
  11. Distinctive:

    When work completes, blocks, fails, or reaches a judgment boundary, report the outcome:

    line 77
  12. Very specific:

    Use --kind completed for finished work, blocked for unresolved blockers, failed for failed attempts, learned for durable learning, and …

    line 88
  13. Very specific:

    Before the final response, check whether code, PR state, merge state, blockers, verification, or durable project knowledge changed since …

    line 94
  14. Distinctive:

    Stop and report clearly if:

    line 98
  15. Very specific:

    Do not continue behind a needs-judgment boundary unless the user explicitly answers it.

    line 108

“Common” is measured: kept by at least 6 in 10 of the 5 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

Continuity Agent Workflow

Continuity is the hosted work graph for coding agents. It is the source of truth for next work, agent packets, blockers, human-judgment boundaries, verification guidance, and work-delta reconciliation.

Use the continuity CLI against the hosted API. Do not ask users for a private source checkout, local graph database, repo-local script, or workspace ID.

Required CLI

If the continuity command is missing, ask the user to install it:

npm install -g @ligence/continuity

Save the workspace token once:

continuity login --api-token <workspace-token>

If the user has a private hosted origin, save it through login:

continuity login --api-url <hosted-api> --api-token <workspace-token>

After login succeeds, rely on ~/.continuity/config.json. Do not tell agents or users to export bearer-token environment variables for normal work.

Start Or Resume

At the start or resume of Continuity-tracked work, check identity, compatibility, readiness, and the next work slice:

continuity whoami --json
continuity version --json
continuity doctor --json
continuity next --agent codex --horizon overnight --compact --json

Use the current agent name instead of codex when running in another agent.

Read the returned sequence or continuation before choosing work. If sequence[0] names runnable work, fetch the packet before editing:

continuity agent-context --node <node-id> --agent codex --json

Follow the packet's authority boundary, stop conditions, verification guidance, and final-answer reconciliation instructions.

Capture Work

Use insert when a user asks for new work, reports a bug, makes a durable decision, or provides an actionable task that should become part of the graph:

continuity insert --text "<work, decision, bug, or goal>" --json

Use intake only when the input should stay in a pending review queue instead of directly mutating graph work:

continuity intake --text "<observation or request>" --json

After inserting graph work, rerun:

continuity next --agent codex --horizon overnight --compact --json

Report Deltas

When work completes, blocks, fails, or reaches a judgment boundary, report the outcome:

continuity report-delta \
  --node <node-id> \
  --kind completed \
  --summary "<what changed>" \
  --evidence "<proof>" \
  --json

Use --kind completed for finished work, blocked for unresolved blockers, failed for failed attempts, learned for durable learning, and needs-judgment when a human decision is required.

A node marked completed can still receive follow-up deltas, evidence, verification updates, merge provenance, or learned context. Do not treat completed status as a reason to skip reconciliation.

Final Response Rule

Before the final response, check whether code, PR state, merge state, blockers, verification, or durable project knowledge changed since the last delta. If a known graph node was touched, run report-delta before answering. If no graph write was made, say why.

Boundaries

Stop and report clearly if:

  • the hosted API is unavailable
  • identity or workspace access fails
  • the graph runtime is not ready
  • selected work requires human judgment
  • verification cannot be made explicit
  • a required external credential, workspace, or billing decision is missing
  • token, budget, or runtime limits prevent safe progress

Do not continue behind a needs-judgment boundary unless the user explicitly answers it. Do not invent work when next returns no runnable sequence; read and report the continuation field.

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.