Maintains persistent bot memory via a structured task ledger and decision log.
npx skills add https://github.com/bkrsna/gemini-cli --skill memory-bkrsna
Please help me install this Agent Skill. Skill: memory Source: https://github.com/bkrsna/gemini-cli/tree/main/tools/gemini-cli-bot/.gemini/skills/memory Command: npx skills add https://github.com/bkrsna/gemini-cli --skill memory-bkrsna
lessons-learned.md with a Task Ledger, Hypothesis Ledger, Decision Log, and investigation findings sections. - Session Synchronization Rituals: Enforces a mandatory Phase 0 context retrieval (reading memory and verifying state via the GitHub CLI) and a Phase 6 memory preservation step at the end of each run. - Memory Pruning: Keeps only the 50 most recent tasks and 20 most recent decision entries to prevent context bloat. - Use Case: A scheduled bot run picks up where the last session ended by reading the Task Ledger, marks merged PRs as DONE, investigates failures, and appends new findings to the Decision Log before finishing. ## Quick Start Activate the memory skill to read lessons-learned.md, synchronize the task ledger with current GitHub state, and record your findings before finishing.High-intent search queries and answers about installing and using this skill.
Store state in a structured Markdown file such as lessons-learned.md containing a task ledger, hypothesis ledger, and decision log. The agent reads this file at the start of each session and updates it before finishing.
Apply rolling-window pruning to the memory file. This Skill keeps only the 50 most recent tasks in the Task Ledger and the 20 most recent entries in the Decision Log to prevent context bloat.
No. Worker agents are strictly forbidden from writing to lessons-learned.md. They return findings to the orchestrator, which remains the sole authority for state preservation.
During Phase 0 it reads the memory file and uses the GitHub CLI (gh pr view or gh issue view) to verify the current state of the trigger, then updates ledger statuses such as marking merged PRs as DONE.
In interactive mode the bot ignores FAILED, STUCK, or pending tasks entirely. Its only goal is to address the specific user comment that triggered the run.
Cogni-AI-OUManage GitHub Issue operations via `gh issue` API and CLI.
kriscendobotOrchestrate multi-seat automated pull request reviews against project rules.
fusebase-devCoordinate parallel AI agent tasks with role-based delegation and verification rules.
oakossManage AI agent tasks and session state via SQLite CLI commands.