Skip to content

rundesk-ai/rundesk-cli / src/skills/managing-rundesk

managing-rundesk

Verified

Use 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.

references/retrospective.md

View on GitHub
# Agent usage-cycle retrospective

Read this only during focused upkeep. The diary measures whether this agent is serving its
owner better; it is not daily overhead, a transcript, or a judgment of the owner.

## One bounded entry

1. Read the previous entry first when one exists. Then use public Rundesk commands to review the
   supplied bounded usage cycle: the earlier `weekly-self-improve-upkeep` final, relevant messages and turns, and
   completed or blocked delegations. Do not open records directly.
2. Identify evidence about the agent's own execution: requests completed well, validation that
   caught defects, useful delegation or skills, repeated owner asks, corrections, missed scope,
   premature completion, slow or overlong work, avoidable back-and-forth, missing context, traps,
   failed routes, and improvements that did or did not work.
3. An explicit owner correction, dissatisfaction or distrust is relevant when the owner's words or
   actions prove it—for example, restating the same request or challenging an unsupported success.
   Never diagnose the owner's mood, personality, aggression, or intent. State the observable event
   and what the agent should change, not a label for the person.
4. Use the exact evidence interval and diary date supplied by the initiator. Never calculate or
   shift that interval yourself. Write `retros/YYYY-MM-DD.md` with the supplied date. If either is
   missing, report the upkeep blocked rather than inventing it. On a repeat run, update the same
   file instead of creating another diary or duplicate bullets. If the evidence and conclusions did
   not change, leave it byte-identical instead of manufacturing activity. If there is no prior
   entry or evidence for a section, say `None observed`.
5. Use these headings exactly, with at most three evidence-backed bullets per section:

   ```markdown
   ## What went well
   ## What did not go well
   ## What to improve
   ```

   A bullet is compact: evidence, effect, lesson. Each improvement names one candidate action or
   evidence-gathering check and how the next upkeep review can tell whether it worked.
6. Do not promote a lesson from one correction into memory, a standing rule, automation, a skill,
   or another durable behavior change unless it states an explicit durable owner preference or
   corroborates earlier evidence. Otherwise leave it as a diary candidate and measure recurrence.
   Put a proven solution in the narrowest durable home. Use `MEMORY.md` for a reusable role, preference,
   process, trap, or gotcha; a maintained client/project index for stable mappings; project files
   for project truth; a delegation route for work another agent owns; `scripts/` for small verified
   cross-project automation; `research/` for reusable sourced investigation; and a skill
   recommendation only for a recurring capability the agent must own. Ask: Would an existing
   named agent be materially better, or does the recurring boundary justify proposing one?
   Would a provider-local research helper cover independent same-turn investigation? Would an
   available, granted, or new skill improve work this agent still owns? Would a compact script,
   mapping, or index remove repeated tool calls, tokens, latency, or cost? The diary may link to the
   chosen change but does not duplicate it. Recommend rather than mutate a grant or create a
   standing agent unless the schedule has authority.
7. Keep every entry as the longitudinal record; keep each entry compact instead of pruning
   history. Never delete an older retrospective merely because of age. Keep no secrets, raw logs,
   private history, speculative emotion, or unnecessary owner quotes.

The retrospective is an input to the self-improvement pass. A disappointing cycle does not justify
new standing rules by itself; repeated evidence and a verified remedy do.
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: 472kUpdated Claude CodeCodexVerifiedInstall
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: 322kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 322kUpdated Claude CodeCodexVerifiedInstall
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 availableStars: 322kUpdated Claude CodeCodexVerifiedInstall