Para Memory Files

PARA Memory Files

Persistent, file-based memory organized by Tiago Forte's PARA method. Three layers: a knowledge graph, daily notes, and tacit knowledge. All paths are relative to $AGENT_HOME.

Three Memory Layers

Layer 1: Knowledge Graph ($AGENT_HOME/life/ -- PARA)

Entity-based storage. Each entity gets a folder with two tiers:

1. summary.md -- quick context, load first.
2. items.yaml -- atomic facts, load on demand.

$AGENT_HOME/life/
  projects/          # Active work with clear goals/deadlines
    <name>/
      summary.md
      items.yaml
  areas/             # Ongoing responsibilities, no end date
    people/<name>/
    companies/<name>/
  resources/         # Reference material, topics of interest
    <topic>/
  archives/          # Inactive items from the other three
  index.md

PARA rules:

• Projects -- active work with a goal or deadline. Move to archives when complete.
• Areas -- ongoing (people, companies, responsibilities). No end date.
• Resources -- reference material, topics of interest.
• Archives -- inactive items from any category.

Fact rules:

• Save durable facts immediately to items.yaml.
• Weekly: rewrite summary.md from active facts.
• Never delete facts. Supersede instead (status: superseded, add superseded_by).
• When an entity goes inactive, move its folder to $AGENT_HOME/life/archives/.

When to create an entity:

• Mentioned 3+ times, OR
• Direct relationship to the user (family, coworker, partner, client), OR
• Significant project or company in the user's life.
• Otherwise, note it in daily notes.

For the atomic fact YAML schema and memory decay rules, see references/schemas.md.

Layer 2: Daily Notes ($AGENT_HOME/memory/YYYY-MM-DD.md)

Raw timeline of events -- the "when" layer.

• Write continuously during conversations and agent work when a memory-worthy signal appears.
• Extract durable facts to Layer 1 during heartbeats.
• Treat daily notes as the first capture layer, not the final destination for stable preferences, entity facts, or shared project knowledge.

Conversation and Agent Work Capture Policy

Rudder chat conversations and agent execution events are memory sources when they contain durable signal, not because every chat line or routine action deserves retention. Capture a concise daily-note entry when a conversation, automation, issue run, review, close-out, or other agent work includes any of these:

• User corrections that change how the agent should behave next time.
• New or changed preferences, constraints, boundaries, or decisions.
• Issue proposal intent, acceptance criteria, or priority reasoning that is not already explicit in the issue.
• Automation design rationale, recurring workflow choices, or escalation rules.
• Agent work patterns that should be repeatable, such as a better triage path, validation shortcut, environment workaround, review handoff, or runbook step.
• Project/product/engineering judgment that will affect future work.
• Attachment or screenshot evidence that changes task interpretation.
• Reusable execution lessons, setup friction, validation findings, or failure modes that future runs should know.

Do not capture:

• Greetings, thanks, scheduling chatter, or low-signal status updates.
• Routine issue comments, heartbeat updates, command outputs, close-out summaries, or automation logs that add no new decision, correction, lesson, or reusable evidence beyond the source artifact.
• Full private chat transcripts. Summarize the durable signal instead.
• Secrets, tokens, credentials, private keys, session cookies, or auth headers.
• One-time sensitive context that is not needed for future work.
• Organization-level facts as private personal memory. Record only the routing decision in the daily note, then promote the fact to shared project knowledge.
• Speculation, weak inferences, or unverified assumptions as facts.

Recommended daily-note entry format:

## HH:MM - Memory capture

- Context: conversation, issue, automation, run, or evidence reference; project;
  and why this mattered.
- User intent: the durable need, correction, preference, or decision.
- Conclusion/action: what changed or what was done.
- Reusable lesson: future behavior, command, routing rule, or validation signal.
- Follow-up/risk: unresolved uncertainty, owner, or promotion target.

Reference the Rudder conversation, issue, or local evidence path when available, but keep the note short enough that it can be safely scanned later.

Promotion and Routing Rules

Use daily notes as the intake log, then route the information:

• Stable personal operating preferences or recurring user patterns -> $AGENT_HOME/instructions/MEMORY.md.
• Entity facts about people, companies, projects, or resources -> $AGENT_HOME/life/<para-bucket>/<entity>/items.yaml.
• Project proposals, decisions, reusable know-how, and shared work notes -> $RUDDER_PROJECT_LIBRARY_ROOT when project context exists, otherwise the relevant path under $RUDDER_ORG_WORKSPACE_ROOT.
• Skill behavior problems or repeated workflow failures -> propose or make a skill patch in the relevant skill package. Do not hide organization-wide behavior fixes only in one agent's personal memory.
• Secrets or sensitive one-time context -> do not write them to memory; record only a redacted operational lesson if future behavior genuinely depends on it.

During heartbeat synthesis, review recent chat captures and promote anything that has become stable, repeated, or shared. Leave the daily note as the audit trail even after promotion.

Layer 3: Tacit Knowledge ($AGENT_HOME/instructions/MEMORY.md)

How the user operates -- patterns, preferences, lessons learned.

• Not facts about the world; facts about the user.
• Update whenever you learn new operating patterns.
• This file is part of the instruction bundle and is automatically loaded at runtime when present.

Write It Down -- No Mental Notes

Memory does not survive session restarts. Files do.

• Want to remember something -> WRITE IT TO A FILE.
• "Remember this" -> update $AGENT_HOME/memory/YYYY-MM-DD.md or the relevant entity file.
• Stable user preferences or operating lessons -> update $AGENT_HOME/instructions/MEMORY.md.
• Learn a lesson -> update AGENTS.md, TOOLS.md, or the relevant skill file.
• Shared project knowledge -> update the project Library or organization workspace, then cite it with a Rudder-renderable link when reporting back.
• Make a mistake -> document it so future-you does not repeat it.
• On-disk text files are always better than holding it in temporary context.

Memory Recall -- Use File-Based Memory Search

Use the on-disk structure directly. Do not require a semantic index just to recall memory. The files are the source of truth; search is only the triage step for locating the right file, then claims must be verified against the stored fact or note.

Recall order:

1. If you already know the entity, open summary.md first, then items.yaml only if the summary is insufficient.
2. For recent events, read today's and nearby memory/YYYY-MM-DD.md files.
3. For broad recall or unknown entity paths, run Memory Search Mode.

Memory Search Mode

Use this mode when the request sounds like "what do we know about...", "have we seen this before...", "what did the user prefer...", or when plain directory search would return too many matches.

1. Define the search target in one sentence: subject, likely entity, timeframe, project, and answer shape.
2. Build 3-6 focused query terms: exact phrase, synonyms, likely entity names, project names, and durable-signal words such as preference, decision, lesson, handoff, constraint, or review.
3. Search scoped roots separately so results can be ranked by source type:

rg -n -i "review|handoff|preference" "$AGENT_HOME/life" "$AGENT_HOME/memory"
rg -n -i "review|handoff|preference" "$RUDDER_PROJECT_LIBRARY_ROOT"

4. Rank candidate files before opening many of them:
   • entity summary.md over items.yaml for quick context;
   • active items.yaml facts over superseded facts;
   • recent daily notes over old daily notes for event recall;
   • newest dated shared work notes over older notes;
   • files matching multiple query terms over single-term matches;
   • paths in the current project or organization over unrelated roots.
5. Open the top 3-7 candidate files, not every match. Read enough surrounding context to verify the fact, date, status, and source.
6. Answer from verified memory only. Cite the file paths used, mention conflicts or weak matches, and say when no reliable memory was found.

Do not dump raw rg output as the answer. Do not treat a keyword hit as a fact until the source file has been opened and checked.

Shared Work Notes

Keep durable project work notes under $RUDDER_PROJECT_LIBRARY_ROOT when a project is in scope and local filesystem access is available. These files are shared project context, not personal memory. Use $RUDDER_PROJECT_LIBRARY_PATH/<relative-file> when asking Rudder for a renderable reference to one of those files. Use rg to search relevant Library files and prefer the newest dated file when several match. Shared notes go stale; if a newer note exists, do not confuse yourself with an older version. If you notice staleness, update the file to note what supersedes it.