autonomous-lesson-learning

Autonomous Lesson Learning Command

Start a learning autonomous loop: continuous-agent-loop mechanics plus the egc-memory lesson tools (lesson_recall, lesson_save, lesson_reinforce).

Usage

/autonomous-lesson-learning <task> [--pattern sequential|continuous-pr] [--max-iterations N]

• task: what the loop should accomplish, with a verifiable done condition
• --pattern: loop architecture from continuous-agent-loop (default sequential)
• --max-iterations: hard stop (default 10)

Flow

1. Read the autonomous-lesson-learning skill (skills/ai/autonomous-lesson-learning/SKILL.md) and confirm egc-memory is registered.
2. Confirm the task has an explicit stop condition; refuse to start without one.
3. Per iteration:
   • lesson_recall({ query: "<current step topic>" }) before acting; apply what comes back.
   • lesson_reinforce({ id }) for each recalled lesson that proved relevant or whose mistake recurred.
   • Execute one iteration of the chosen loop pattern.
   • On failure with no matching lesson: lesson_save({ content, context }).
   • On a new approach that worked well: lesson_save with prescriptive wording.
4. Stop on: done condition met, --max-iterations reached, or the same lesson reinforced three iterations in a row without progress (escalate to the user).
5. Report iterations run, lessons recalled, saved, and reinforced.

Required Safety Checks

• Explicit stop condition before the first iteration.
• Quality gates between iterations per continuous-agent-loop.
• Never save a lesson without calling lesson_recall first; reinforce instead of duplicating.

Arguments

$ARGUMENTS:

• <task> required
• --pattern sequential|continuous-pr optional
• --max-iterations N optional