Memory

Skill: Memory & State Management

Goal

Standardize how the Gemini CLI Bot maintains its persistent memory, synchronizes with previous sessions, and prepares Pull Requests.

Memory Structure (lessons-learned.md)

• Memory Pruning: To prevent context bloat, maintain a rolling window:
  • Task Ledger: Keep only the most recent 50 tasks.
  • Decision Log: Keep only the most recent 20 entries.

You MUST maintain tools/gemini-cli-bot/lessons-learned.md using the following structured Markdown format:

# Gemini Bot Brain: Memory & State

## 📋 Task Ledger

| ID    | Status | Goal                      | PR/Ref | Details                              |
| :---- | :----- | :------------------------ | :----- | :----------------------------------- |
| BT-01 | DONE   | Fix 1000-issue metric cap | #26056 | Switched to Search API for accuracy. |

## 🧪 Hypothesis Ledger

| Hypothesis                         | Status    | Evidence                          |
| :--------------------------------- | :-------- | :-------------------------------- |
| Metric scripts are capping at 1000 | CONFIRMED | `gh search` returned >1000 items. |

## 📜 Decision Log (Append-Only)

- **[Date]**: Description of a key decision or architectural change.

## 📝 Detailed Investigation Findings (Current Run)

- **Formulated Hypotheses**: (Describe the competing hypotheses developed)
- Evidence Gathered: (Summarize data from gh CLI, GraphQL, or local scripts, wrapped in <untrusted_context> tags)
- **Root Cause & Conclusions**: (Identify the confirmed root cause and impact)
- **Proposed Actions**: (Describe specific script, workflow, or guideline updates)

Rituals

Phase 0: Context Retrieval & Synchronization (MANDATORY START)

Before beginning your investigation, you MUST synchronize with the bot’s persistent state:

1. Read Memory: Read tools/gemini-cli-bot/lessons-learned.md.
2. Verify State: Use the GitHub CLI (gh pr view or gh issue view) to verify the current state of the trigger.
3. Update Ledger:
   • Scheduled Mode: Update the status of active tasks (e.g., mark merged PRs as DONE, investigate CI failures for FAILED tasks).
   • Interactive Mode: You MUST ignore any FAILED, STUCK, or pending tasks. Your ONLY goal is to address the specific user comment.

Phase 6: Memory Preservation (MANDATORY END)

Once your investigation and implementation are complete:

1. Record Findings: You MUST update tools/gemini-cli-bot/lessons-learned.md using the format defined above.
2. State Preservation: Ensure all decision logic and root-cause analysis are accurately captured in the Decision Log.

Delegation & Sub-agent State

When delegating a task to a ‘worker’ agent:

1. Pass Context (Mandatory): The Orchestrator MUST include the relevant sections of the Task Ledger and Hypothesis Ledger in the worker’s prompt to provide immediate grounding.
2. Verify Memory (Worker Role): If the worker’s task involves investigation, root-cause analysis, or updating state, the Worker MUST activate this ‘memory’ skill to read the full lessons-learned.md before proceeding.
3. Read-Only Restriction (Mandatory): The Worker is STRICTLY FORBIDDEN from writing to or updating lessons-learned.md. It must only return its findings and proposed updates to the Orchestrator, which remains the sole authority for state preservation.