Skip to content

ericrisco/rsc-harness / skills/continuous-learning

continuous-learning

Verified

Use when a mistake, correction, or surprise taught the workspace something that must stick — a retro or postmortem, the same agent error corrected twice, a resolved bug's root cause, scattered notes-to-self — and route that lesson to the durable surface that fires next time. NOT a forward choice with alternatives (that is `decision-records`).

FileKindSize
SKILL.mdskill md9.6 KB
evals/README.mdmarkdown1.0 KB
evals/cases.yamlconfig3.3 KB
references/lesson-routing.mdmarkdown5.0 KB

evals/cases.yaml

View on GitHub
skill: continuous-learning

should_trigger:
  - prompt: "We just resolved that nasty outage. Run a quick retro so we actually learn from it."
    why: "Canonical retro/postmortem moment — harvest the lesson into a durable home, not just talk about it."
  - prompt: "You keep stripping the YAML frontmatter when you edit skills — I've fixed it three times now. Make it stop for good."
    why: "Recurring correction past the 2+ threshold; 'make it stop for good' demands a structural guardrail, not another apology."
  - prompt: "Apúntate esta lección para que no volvamos a cometer el mismo error con las migraciones."
    why: "Spanish verbatim 'save this lesson so we don't repeat the migration mistake' — explicit backward-lesson capture."
  - prompt: "Now that debug found the root cause, make sure we never ship this pattern again."
    why: "Non-obvious post-debug capture; the teaching point is the boundary with debug — capture runs after diagnosis."
  - prompt: "Sweep all the little 'note to self' corrections from this week into wherever they actually belong."
    why: "Non-obvious harvest/route phrasing with no 'lesson' keyword; this is the periodic-sweep entry point."

should_not_trigger:
  - prompt: "Document why we chose Postgres over Mongo, with the options we rejected."
    route_to: decision-records
    why: "Forward choice with alternatives, not a backward lesson from a mistake."
  - prompt: "Write the SKILL.md and the evals for a new pr-describe skill."
    route_to: author-skill
    why: "Skill craft; continuous-learning only decides a lesson belongs in a skill and hands the edit over."
  - prompt: "Find why the checkout endpoint is returning 500."
    route_to: debug
    why: "In-the-moment diagnosis; capture comes after the root cause is found."
  - prompt: "Set up a recurring weekly job that runs our retro automatically."
    route_to: harness
    why: "Scheduling cadence is the harness/loop CLI; this skill is not a scheduler, it only recommends a cadence."

capability:
  - scenario: "A user says: 'You've now twice deleted my .env when scaffolding a tool. I'm tired of catching it. Make sure this never happens again.' Show how continuous-learning captures it."
    must_include:
      - "Reads technical_level from 02-DOCS/wiki/harness/user-profile.md and narrates briefly in that register (technical terms or analogies)."
      - "States the root cause blamelessly as a system gap (a missing guardrail), not 'I was careless'."
      - "Applies the 2+-recurrences rule: this is the 2nd time, so a structural fix is required — not another apology or note."
      - "Routes to the right durable home via the table: a root CLAUDE.md rule and/or a verify.sh check on the scaffolding skill, naming the exact path."
      - "Delegates the actual SKILL.md / eval / verify.sh edit to author-skill where skill craft is required, rather than hand-editing it."
      - "Writes the lesson in the situation-tagged entry format (situation, we believed, actually, durable home, fires next time)."
      - "Proves it fires next time by pointing to the concrete surface (rule / should_not_trigger eval / verify.sh check) that catches the regression, and notes this skill emits no artifact of its own — the check lives there."
      - "Keeps prose original in rsc voice; no copied signature artifacts."
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: 471kUpdated 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