agentleFS
conorbronsdonconorbronsdon/agent-context-os/.agents/skills/context-update/SKILL.md

context-update

Skill30 starsChanged 37 days agoMITView on GitHub

Save a brief mid-session checkpoint to this workspace and update current state only when a priority or open thread changed. Use only when the user explicitly asks to checkpoint or save current progress.

The file

We split this skill into the rules it gives an agent

8 rules. About 0 are common, 4 are distinctive, and 4 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈4

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 4

    Very specific

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

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

    When the host executes commands inside Bash, replace the bash executable shown below with "$BASH" to reuse that shell.

    line 12
  2. Very specific:

    Use the exact roots supplied by the host attachment: KernelRoot is the trusted Context OS product containing scripts/contextos.sh.

    line 17
  3. Distinctive:

    Do not search upward or infer a root from cwd or the skill installation.

    line 26
  4. Very specific:

    Create a reviewed JSON payload under <ContextRoot>/.context-os/inputs/ with progress as one to three factual strings.

    line 45
  5. Distinctive:

    Run exactly one matching form:

    line 51
  6. Distinctive:

    Wait for explicit approval of that exact proposal.

    line 62
  7. Distinctive:

    If any target changed after proposal creation, create and review a new proposal.

    line 69
  8. Very specific:

    If a durable decision was made, mention that $end can add it to the decision log.

    line 71

≈ “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

Checkpoint a workspace session

Execution roots (required)

When the host executes commands inside Bash, replace the bash executable shown below with "$BASH" to reuse that shell. This avoids selecting a different Bash from PATH, such as the Windows WSL launcher inside a Git Bash session. In a terminal outside Bash, select the installed Bash explicitly.

Use the exact roots supplied by the host attachment: KernelRoot is the trusted Context OS product containing scripts/contextos.sh; ContextRoot owns tracked identity and lifecycle state; and WorkingRoot is the ordinary application. For an external attachment, require all three exact absolute paths and run:

bash <KernelRoot>/scripts/contextos.sh --context-root <ContextRoot> --working-root <WorkingRoot> <command>

Do not search upward or infer a root from cwd or the skill installation. The kernel must validate the ignored local binding before strict lifecycle work. A missing, moved, stale, linked, nested, or mismatched binding stops the workflow; use the explicit project rebind proposal after a legitimate move. ContextRoot owns all lifecycle writes. WorkingRoot is read-only evidence. The colocated bash scripts/contextos.sh <command> compatibility form remains valid.

Throughout this procedure, resolve every context and local-artifact path beneath ContextRoot. In split mode, spell local paths as absolute <ContextRoot>/... paths and invoke every lifecycle command through the absolute KernelRoot wrapper with both exact role options. In colocated mode, run the relative compatibility commands from the colocated root.

Save continuity with minimal churn through the deterministic lifecycle kernel.

Procedure

  1. Identify what was completed, any decisions, and whether a priority or open thread changed. Do not invent progress.
  2. Create a reviewed JSON payload under <ContextRoot>/.context-os/inputs/ with progress as one to three factual strings. Make each progress entry one self-contained sentence naming its subject and stating one fact; avoid label-prefixed fragments and multiple facts in one sentence, but keep a status and its qualifiers in the same sentence. Apply the same rule to handoff prose. Include current_markdown only when a priority shifted, a thread opened or closed, or a tracked task completed. When present, it is the complete desired ContextRoot current.md before the kernel advances its date and history. Preserve unrelated content and ordering.
  3. Run exactly one matching form:

    split:     bash <KernelRoot>/scripts/contextos.sh --context-root <ContextRoot> --working-root <WorkingRoot> propose update --input <ContextRoot>/.context-os/inputs/<payload.json>
    colocated: bash scripts/contextos.sh propose update --input .context-os/inputs/<payload.json>

    The kernel owns session append behavior, dates, the single Last Updated line, and same-day history. Present every returned diff and the exact proposal digest. The digest binds the exact content but does not authenticate a human approver; rely on the host permission boundary for confirmation.

  4. Wait for explicit approval of that exact proposal. Then run:

    split:     bash <KernelRoot>/scripts/contextos.sh --context-root <ContextRoot> --working-root <WorkingRoot> apply <ContextRoot>/<proposal> --confirm <digest> --runtime <active-runtime>
    colocated: bash scripts/contextos.sh apply <proposal> --confirm <digest> --runtime <active-runtime>

    If any target changed after proposal creation, create and review a new proposal. Never bypass stale-write, path-containment, or locking checks.

  5. If a durable decision was made, mention that $end can add it to the decision log; do not expand a checkpoint into a full close workflow.
  6. Report the receipt path and changed files. Do not edit lifecycle state directly, commit, push, or touch files only to refresh timestamps.

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.