agentleFS
Sign inStart free
nitekeepernitekeeper/atelier/skills/load/SKILL.md

load

Skill0 starsChanged 5 months agoMITView on GitHub

Use when resuming work mid-arc and needing the latest session context loaded into the conversation.

8 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

17 rules. 0 are common, 8 are distinctive, and 9 are very specific to this project.

  • 0

    Common

    Most of the 6 edited versions we read keep these.

  • 8

    Distinctive

    This author's own choices. Worth a look.

  • 9

    Very specific

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

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

    Call load at the start of every session, before any other action.

    line 11
  2. Very specific:

    Settings recommendation (after pre-flight): if startup_check() returned a settings_rec_offer with eligible=True and non-empty changes, read …

    line 29
  3. Distinctive:

    BEFORE proceeding to the rest of this skill.

    line 32
  4. Distinctive:

    Run (from the target project root):

    line 38
  5. Distinctive:

    If exactly one project exists: use it automatically.

    line 44
  6. Distinctive:

    If multiple exist: display the list (id, name, phase) and ask the user: "Which project are you resuming?" Wait for a response, then …

    line 45
  7. Very specific:

    If none exist: announce "No projects found.

    line 46
  8. Distinctive:

    If a session row was returned, announce:

    line 56
  9. Very specific:

    If closed_at is null, omit the "Last closed" clause:

    line 59
  10. Very specific:

    If blocking_reason is non-null, append:

    line 62
  11. Distinctive:

    If the command returns "No session found for this project.", announce:

    line 65
  12. Very specific:

    Run two Memex ask queries:

    line 69
  13. Very specific:

    Do not begin executing next_action automatically.

    line 75
  14. Distinctive:

    Always run steps 1–4 before doing any work.

    line 81
  15. Very specific:

    Never skip the project-identification step — read-latest requires a valid project_id.

    line 82
  16. Very specific:

    Never begin executing next_action without user confirmation.

    line 84
  17. Distinctive:

    If project identification or session read fails, surface the error to the user before proceeding.

    line 85

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

load

Session-open command. Reads the most recent session from the DB and surfaces relevant knowledge from Memex before work begins.

When to use

Call load at the start of every session, before any other action.

Pre-flight (always first)

Run from scripts.atelier_entrypoint import startup_check; startup_check().

Branch on the returned action:

  • proceed-local — Memex is not installed. Continue with the rest of this skill's recipe; all writes go to the project-local .ai/atelier.db.
  • proceed-memex — Memex is installed and bootstrapped. Continue; all writes go through Memex.
  • prompt-migration — Memex is installed but this project still has a local DB. Read internal/migrate-local-to-memex/SKILL.md and follow its prompt protocol. After the user answers, restart the pre-flight (startup_check() will now return proceed-memex or proceed-local depending on the user's choice).

Settings recommendation (after pre-flight): if startup_check() returned a settings_rec_offer with eligible=True and non-empty changes, read internal/settings-recommendation/SKILL.md and follow its prompt protocol BEFORE proceeding to the rest of this skill. After the user's choice, continue the original command.

Procedure

  1. Identify the active project. Run (from the target project root):

    python3 atelier/scripts/projects.py list

    (DB: .ai/memex.db)

    • If exactly one project exists: use it automatically. Announce: "Found project: [name] (ID [id], phase: [phase])."
    • If multiple exist: display the list (id, name, phase) and ask the user: "Which project are you resuming?" Wait for a response, then confirm: "Resuming [name] (ID [id], phase: [phase]). Correct? (yes/no)"
    • If none exist: announce "No projects found. Run internal/project/SKILL.md (create) to set up a project first." Stop.
  2. Read the latest session. Run:
    python3 atelier/scripts/session.py read-latest <project_id>
    (DB: .ai/memex.db)

  3. Announce session state. If a session row was returned, announce:

    "Resuming: [current_tasks]. Status: [status]. Last closed: [closed_at]. Next action: [next_action]."

    If closed_at is null, omit the "Last closed" clause:

    "Resuming: [current_tasks]. Status: [status]. Next action: [next_action]."

    If blocking_reason is non-null, append:

    "Blocked: [blocking_reason]."

    If the command returns "No session found for this project.", announce:

    "No previous session found for this project. Starting fresh."

  4. Surface Memex context (only if a session row was returned). Run two Memex ask queries:
    • Query 1: the value of current_tasks from the session record
    • Query 2: the value of next_action from the session record (or ask the user which files/area to query if next_action is not file-specific) Present the results. If Memex returns nothing relevant, say so briefly and proceed.
  5. Confirm next action with the user. Do not begin executing next_action automatically. Ask:

    "Ready to continue from: '[next_action]'? Or do you want to start somewhere else?" Begin work only after the user confirms or redirects.

Hard rules

  • Always run steps 1–4 before doing any work.
  • Never skip the project-identification step — read-latest requires a valid project_id.
  • Both projects.py and session.py use .ai/memex.db. Pass no db path; the default is correct.
  • Never begin executing next_action without user confirmation.
  • If project identification or session read fails, surface the error to the user before proceeding.

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