memory-recall

Memory Recall

Read memory.json and surface the most relevant entries for the current task.

Step 1: Load memory file

Check if memory.json exists at the project root:

• If it does not exist, suggest running /memory-init first and stop.
• If it exists, read its full contents.

Step 2: Understand the current task

Before filtering memory, understand what the user is about to work on:

• Read any task description provided by the user.
• Note key terms: file names, feature names, component names, issue IDs.
• If no task is specified, return the full recent session history (last 3 sessions).

Step 3: Surface relevant entries

Filter and present memory entries relevant to the current task:

Project context (always include)

• Project name, description, stack
• Any stack-specific conventions from the architecture log

Architecture decisions (filter by relevance)

• Include decisions that mention the same component, feature, or technology as the task
• Format: [date] Decision: <decision> (rationale: <rationale>)

Known issues (filter by relevance)

• Include open issues related to the task area
• Mark resolved issues as [resolved] and include only if directly relevant
• Format: [id] <description> -- status: <open|resolved>

Recent sessions (always include last 2)

• Date, summary, and what was noted as "next"
• Highlight if a previous session's "next" matches the current task

Step 4: Format the context block

Present recalled memory as a concise context block:

--- Memory Context ---
Project: <name> | Stack: <stack>

Architecture decisions relevant to this task:
- [date] <decision>

Known issues:
- [id] <description> [open]

Recent sessions:
- [date] <summary> | Next: <next>
- [date] <summary> | Next: <next>
--- End Memory ---

Step 5: Note gaps

If the task involves areas with no memory entries, note:

• "No prior decisions recorded for "
• "No known issues in "
• Suggest updating memory after the session with /memory-update