rundesk-ai/rundesk-cli / src/skills/managing-rundesk
managing-rundesk
VerifiedUse when managing this agent, any other agent in its Rundesk install, or the install itself. It supplies verified operational workflows and safety rules for agent instructions, gateways, conversations, channels, schedules, providers, skills, backups, configuration, credentials, and the Rundesk lifecycle. Do not use for ordinary project work or for deciding, performing, or reviewing delegation.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 5.0 KB |
| references/agent-instructions.md | markdown | 12.3 KB |
| references/agents.md | markdown | 5.9 KB |
| references/backups.md | markdown | 2.1 KB |
| references/channels.md | markdown | 6.6 KB |
| references/configuration.md | markdown | 3.6 KB |
| references/conversations.md | markdown | 5.3 KB |
| references/gateways.md | markdown | 3.1 KB |
| references/lifecycle.md | markdown | 3.7 KB |
| references/maintenance.md | markdown | 5.9 KB |
| references/oauth-login.md | markdown | 3.4 KB |
| references/permissions.md | markdown | 3.6 KB |
| references/providers.md | markdown | 2.8 KB |
| references/retrospective.md | markdown | 3.8 KB |
| references/schedules.md | markdown | 6.4 KB |
| references/self-improvement.md | markdown | 10.7 KB |
| references/skills.md | markdown | 6.6 KB |
| references/teams.md | markdown | 3.2 KB |
references/maintenance.md
View on GitHub# Agent continuity and workspace maintenance
Read this only when maintenance itself is the task. Ordinary work records newly useful durable
context and removes its own scratch; it does not inventory the whole home.
## Focused pass
1. Read `MEMORY.md` and only the home indexes it links. Inventory home loose files and scratch
directories without following symlinks. Reuse the generated `plans/`, `research/`, `scripts/`,
`retros/`, and `tasks/` areas for agent-owned work; keep each README and follow its boundary.
For a symlink, inspect only link metadata and its target spelling; never open the link or target.
The one exception is a held skill's `SKILL.md` and the references it requires. Before opening or
natively loading each such file, check path metadata only, never contents:
- `"$RUNDESK_COMMAND" skills list "$RUNDESK_AGENT"` lists `<name>` with a FROM catalog, not `—`;
- for FROM `<catalog>`, the `realpath` of the exact path you will open lies inside
`<library>/<catalog>/`, where `<library>` is the `realpath` of the directory on the first line
of `"$RUNDESK_COMMAND" skills`;
- for FROM `<catalog> (--as)`, `skills/<name>` is a real directory, not a symlink, and the file's
`realpath` lies inside that directory's `realpath`.
```sh
library=$(realpath "$("$RUNDESK_COMMAND" skills | sed -n '1s/^skills in //p')") &&
file=$(realpath "skills/managing-rundesk/references/retrospective.md") &&
case "$file" in "$library/rundesk/"*) echo inside ;; *) echo forbidden ;; esac
```
A file that fails, is missing, or does not resolve stays a forbidden link target. Never change,
replace, or remove a link or its target.
Preserve and list `retros/`, but do not open its entries; the next phase owns that evidence.
2. Keep compact owner preferences, durable role and responsibilities, reusable cross-project
process and gotchas, and small active-project pointers: stable location, purpose, role, and
authoritative overview.
3. Correct or merge duplicates. Remove a mapping only when it is explicitly marked retired/stale
or authoritative context confirms that; retain an unavailable active mapping and do not add
availability status. Remove loops explicitly marked completed/delivered.
4. Remove commands or deliverable paths, working or draft paths, changing project status, report
formatting, supersession history, and task-only methods from personal memory. Keep lasting truth
in the project; keep only active resumption state in `tasks/` and remove its brief when the task
closes. Preserve a still-open owner commitment until it is closed or safely moved; when several
active commitments outgrow memory, maintain one canonical `OPEN_ITEMS.md` instead of dropping
them merely because they have dates.
An earned shared index may keep a stable entrypoint that avoids repeated discovery. Use one
purpose-named index such as `PROJECTS.md`, `CLIENTS.md`, or `OPEN_ITEMS.md` only when several
entries earn it; never one home note per project or all three as boilerplate.
5. Remove each current-task temporary path wherever it is, plus confirmed agent-created obsolete
clutter. Remove an agent-created scratch directory when left empty. Preserve deliverables,
project or user files, files of uncertain ownership or value, provider-managed directories,
symlinks, and their targets.
Move an agent-owned durable item into the matching generated area only after proving ownership,
purpose, and that no canonical path or link depends on its current location. Do not reorganize
uncertain material for appearance.
6. Reread every changed index and verify each removal and retained sentinel. During combined upkeep,
keep that evidence for the later phases and the short final required by the combined contract.
When maintenance is the whole task, report what changed, was preserved, remains uncertain, and
how it was checked.
Never infer that age, a missing mount, or an unavailable path makes a file or mapping stale. Never
move project state into home merely to organize it. If safe ownership or obsolescence cannot be
proved, preserve the item and report it.
## Tidy versus cluttered
A tidy home is small, canonical, and useful rather than empty:
```text
home/
MEMORY.md compact first-read map; links to PROJECTS.md
PROJECTS.md one maintained index earned by the agent's role
plans/ resumable cross-project agent plans
research/ reusable sourced findings
scripts/ verified agent-owned automation
retros/ dated self-improvement history
tasks/ active resumable briefs, removed when closed
```
Project code, current status, and project decisions remain in each external project. A role that
does not need `PROJECTS.md` should not create it merely to match the example.
A cluttered home has signals worth investigating:
```text
home/
MEMORY.md long history, closed loops, changing status
PROJECTS.md
projects-new.md parallel index with overlapping truth
acorn-copy/ copied project checkout
report.md
report-final.md
report-final-2.md competing deliverables with no canonical location
notes.md detached context nothing links
temp-output.txt
scratch/old-result.txt abandoned intermediates
```
Similarity to this tree is not deletion authority. Establish which file is canonical, who owns
each item, whether it is still needed, and where project state belongs. Preserve anything uncertain.
## Combined upkeep contract
Use one focused schedule for maintenance, retrospective, and self-improvement rather than competing
sweeps. Workspace and continuity maintenance runs first; the retrospective records the supplied usage cycle;
the evidence review then starts from clean, verified continuity. The strict phase order and finish
gate are in the [self-improvement reference](self-improvement.md#combined-upkeep-contract).
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |