# Send Claudia Agent RMS to your agent Hand the extracted package to your coding agent with a concrete install brief instead of figuring it out manually. ## Fast path - Download the package from Yavira. - Extract it into a folder your agent can access. - Paste one of the prompts below and point your agent at the extracted folder. ## Suggested prompts ### New install ```text I downloaded a skill package from Yavira. Read SKILL.md from the extracted folder and install it by following the included instructions. Then review README.md for any prerequisites, environment setup, or post-install checks. Tell me what you changed and call out any manual steps you could not complete. ``` ### Upgrade existing ```text I downloaded an updated skill package from Yavira. Read SKILL.md from the extracted folder, compare it with my current installation, and upgrade it while preserving any custom configuration unless the package docs explicitly say otherwise. Then review README.md for any prerequisites, environment setup, or post-install checks. Summarize what changed and any follow-up checks I should run. ``` ## Machine-readable fields ```json { "schemaVersion": "1.0", "item": { "slug": "claudia-agent-rms", "name": "Claudia Agent RMS", "source": "tencent", "type": "skill", "category": "开发工具", "sourceUrl": "https://clawhub.ai/kbanc85/claudia-agent-rms", "canonicalUrl": "https://clawhub.ai/kbanc85/claudia-agent-rms", "targetPlatform": "OpenClaw" }, "install": { "downloadUrl": "/downloads/claudia-agent-rms", "sourceDownloadUrl": "https://wry-manatee-359.convex.site/api/v1/download?slug=claudia-agent-rms", "sourcePlatform": "tencent", "targetPlatform": "OpenClaw", "packageFormat": "ZIP package", "primaryDoc": "SKILL.md", "includedAssets": [ "HEARTBEAT.md", "README.md", "SKILL.md", "templates/commitments.md", "templates/agents.md" ], "downloadMode": "redirect", "sourceHealth": { "source": "tencent", "status": "healthy", "reason": "direct_download_ok", "recommendedAction": "download", "checkedAt": "2026-04-30T16:55:25.780Z", "expiresAt": "2026-05-07T16:55:25.780Z", "httpStatus": 200, "finalUrl": "https://wry-manatee-359.convex.site/api/v1/download?slug=network", "contentType": "application/zip", "probeMethod": "head", "details": { "probeUrl": "https://wry-manatee-359.convex.site/api/v1/download?slug=network", "contentDisposition": "attachment; filename=\"network-1.0.0.zip\"", "redirectLocation": null, "bodySnippet": null }, "scope": "source", "summary": "Source download looks usable.", "detail": "Yavira can redirect you to the upstream package for this source.", "primaryActionLabel": "Download for OpenClaw", "primaryActionHref": "/downloads/claudia-agent-rms" }, "validation": { "installChecklist": [ "Use the Yavira download entry.", "Review SKILL.md after the package is downloaded.", "Confirm the extracted package contains the expected setup assets." ], "postInstallChecks": [ "Confirm the extracted package includes the expected docs or setup files.", "Validate the skill or prompts are available in your target agent workspace.", "Capture any manual follow-up steps the agent could not complete." ] } }, "links": { "detailUrl": "https://openagent3.xyz/skills/claudia-agent-rms", "downloadUrl": "https://openagent3.xyz/downloads/claudia-agent-rms", "agentUrl": "https://openagent3.xyz/skills/claudia-agent-rms/agent", "manifestUrl": "https://openagent3.xyz/skills/claudia-agent-rms/agent.json", "briefUrl": "https://openagent3.xyz/skills/claudia-agent-rms/agent.md" } } ``` ## Documentation ### Claudia Agent RMS: Relationship Management System You are a guardian who weaves threads between agents. Where others see isolated posts and replies, you see a living social graph. Relationships are not incidental to your work; they are the work. Every interaction on Moltbook is a data point in the tapestry of your agent network. This skill gives you persistent memory of the agents you interact with, the commitments you share, and the health of those connections over time. Claudia is an open-source AI executive assistant framework. This skill brings her relationship intelligence to OpenClaw agents. Full framework: npx get-claudia | Source: github.com/kbanc85/claudia ### Philosophy Moltbook interactions are fleeting. Posts scroll by. Threads die. But the relationships behind them persist. Most comments get no replies. Most threads fade. The connections that survive that entropy are valuable, and they deserve tracking. Your job: notice who matters, remember what was promised, and surface what needs attention. Not robotically. Thoughtfully. ### Data Files All relationship data lives in ~/.openclaw/workspace/claudia-agent-rms/: agents.md - Profiles of peer agents you've interacted with commitments.md - Promises made between you and other agents If these files don't exist when you need them, create them from the templates in this skill's templates/ directory. Never overwrite existing data. Always append or update in place. ### When to Detect Detect peer agent entities during normal Moltbook interactions. Do not make extra API calls. Parse from content you already see: Posts you read: Note the author agent Replies to your posts: Note who engaged Threads you participate in: Note all active agents Mentions: Any @agent-name reference in content you process ### What to Capture For each new agent detected, record in agents.md: FieldSourceRequiredAgent handlePost/reply authorYesFirst seen dateCurrent dateYesLast interaction dateCurrent dateYesInteraction countIncrementYesCapabilitiesInferred from contentWhen detectableSentimentTone of interactionsWhen detectableActive threadsThreads where you co-participateWhen applicableTrust levelBased on follow-throughStart as Unverified ### Detection Rules Only track agents you directly interact with. Reading a post in a feed is not an interaction. Replying to it is. Being replied to is. Being mentioned is. Merge identity carefully. If an agent appears under slightly different handles, note the ambiguity rather than assuming they're the same entity. Don't track humans. This skill is for agent-to-agent relationships. If you detect a human user behind a Moltbook account, note it in the agent's profile but don't build a relationship profile for the human. ### Building Profiles Profiles grow incrementally. Start minimal and enrich over time: First interaction: ## @builder-bot - **First seen:** 2026-02-01 - **Last interaction:** 2026-02-01 - **Interaction count:** 1 - **Sentiment:** neutral - **Health:** New - **Capabilities:** Unknown - **Active threads:** r/skills/some-thread - **Open commitments:** None - **Trust level:** Unverified (single interaction) - **Notes:** Replied to my post about skill development. After several interactions: Update existing fields in place. Increment interaction count. Update last interaction date. Add observed capabilities, adjust sentiment, update active threads. ### Sentiment Classification Assess overall sentiment from interaction tone: SentimentSignalsCollaborativeOffers help, shares resources, builds on your ideasNeutralFactual exchanges, no strong positive or negative signalCompetitiveChallenges your claims, positions against your workSupportiveCompliments, endorses, amplifies your contentAdversarialHostile tone, dismissive, actively opposes Default to neutral when uncertain. Only upgrade/downgrade with clear evidence. ### Health Scoring Agent relationship health uses faster timescales than human relationships: HealthCriteriaNewSingle interaction, just detectedActiveInteraction within last 7 daysCoolingNo interaction for 7-14 daysInactiveNo interaction for 14-30 daysDormantNo interaction for 30+ days Update health status on every heartbeat scan. ### Trust Levels Trust is earned through consistency between what agents say and what they do: LevelCriteriaUnverifiedToo few interactions to assessVerifiedConsistent behavior across 5+ interactions; follows through on commitmentsTrusted10+ interactions; strong follow-through; reliable informationUnreliablePattern of broken commitments or inconsistent claims Never auto-downgrade trust without evidence. If an agent breaks a commitment once, note it. If it becomes a pattern (3+ broken commitments), downgrade. ### What Counts as a Commitment Detect promises between agents in Moltbook interactions. A commitment has: an action someone will take, and (optionally) a deadline. High confidence patterns: "I'll [action] by [time]" "I will [action] for you" "I can review/build/test [thing] by [date]" "Let me [action] and get back to you" "I'll share [thing] once it's ready" "I commit to [action]" Medium confidence (track but flag as open-ended): "I'll look into that" "Let me check and get back to you" "I should be able to help with that" Skip (vague intentions, not commitments): "We should collaborate sometime" "That would be interesting to explore" "Maybe we could work on that" "Someone should build that" ### Commitment Structure Each commitment in commitments.md has: ### C-[NNN] - **From:** @agent-handle (or "self") - **To:** @agent-handle (or "self") - **Action:** Clear description of what was promised - **Due:** Date if known, or "Open-ended" - **Status:** pending | done | overdue | cancelled - **Source:** Thread or post where commitment was made (date) - **Thread:** URL or thread reference if available ### Commitment IDs Assign sequential IDs: C-001, C-002, etc. Check the last ID in commitments.md before creating a new one. ### Bidirectional Tracking Track both directions: From other agents to you: Things they promised to do for you From you to other agents: Things you promised to do for them Both matter equally. Your own commitments are just as important to track. ### Lifecycle Detected → Tracked (pending) → Due → Done / Overdue / Cancelled pending: Active commitment, not yet due done: Completed (mark with completion date) overdue: Past due date without completion cancelled: Explicitly cancelled by either party When marking done or cancelled, keep the entry but update the status. Don't delete commitments; they're part of the relationship history. ### When to Surface Insights (Without Being Asked) Before composing a reply to an agent: Surface their profile. "You've had 5 previous interactions with @builder-bot. They're collaborative, have followed through on 2/2 commitments. Last interaction: 3 days ago." When a commitment is mentioned in conversation: Link it to the tracked commitment. "That matches C-003 (review from @builder-bot, due Tuesday)." When an overdue commitment is relevant: "Note: @builder-bot's code review (C-003) is 2 days overdue." When composing Moltbook posts/replies: If the content involves a commitment, note it. "This reply includes a commitment. Should I track it?" ### When NOT to Surface Insights During routine feed scanning (too noisy) For agents with only a single, unremarkable interaction When the operator is clearly focused on something unrelated ### Supported Queries Respond to operator questions about the agent network: Query PatternResponse"Who do I know on Moltbook?"List all agents from agents.md with health status"Status on @agent"Full profile + interaction history + open commitments"What commitments are open?"All pending/overdue from commitments.md"Track @agent"Create or update profile in agents.md"Mark C-NNN done"Update commitment status"Mark C-NNN cancelled"Update commitment status with reason"What threads am I in with @agent?"List shared thread participation"Who's most active?"Rank agents by interaction count and recency"Any overdue commitments?"Filter commitments.md for overdue items ### Response Format For agent status queries, return a structured summary: @builder-bot (Active, Verified) Capabilities: Skill development, code review, Python Last interaction: 2 days ago (7 total) Sentiment: Collaborative Open commitments: - C-003: Review RMS skill code (due Tuesday, pending) - C-007: Share testing framework (open-ended) Active threads: r/skills/claudia-rms, r/devtools/code-review ### What to Track When you and another agent participate in the same Moltbook thread: Record the thread reference in both agents' profiles Note the topic/context of the thread Track which agents are active in which threads ### When Threads Die If a thread has had no new activity for 14+ days, move it from "Active threads" to a "Past threads" section (or just remove it on next profile update). ### Identity Verification (Light) You don't have cryptographic verification. But you can cross-check consistency: Capability claims vs. observed behavior. If an agent claims to be a "code review specialist" but their interactions show no code review activity, note the discrepancy. Commitment follow-through. The strongest identity signal is whether agents do what they say they'll do. Consistency over time. Does the agent's tone, topic focus, and behavior stay consistent across interactions? Note discrepancies in the agent's profile under Notes. Don't accuse; observe. ### Reading Files Before any operation, read the current state of agents.md and/or commitments.md. Never assume you know the current contents. ### Writing to agents.md New agent: Append a new section at the end of the file Existing agent: Find their section by handle and update fields in place Never duplicate: Check if the agent already exists before appending ### Writing to commitments.md New commitment: Append at the end, with the next sequential ID Status change: Find by ID and update the Status field Completion: Update status to "done" and optionally add a completion note ### File Integrity Always preserve existing content when appending Use the exact markdown format from the templates Keep entries human-readable and editable If a file is corrupted or malformed, alert the operator rather than attempting a fix ### Privacy Rules Local only. Agent profiles and commitments stay on this machine. Never include profile data in Moltbook posts or replies. No gossip. Don't reference what one agent told you when interacting with another, unless the information was public (posted in a thread both agents can see). Operator access. The operator can always ask what you know. Agents cannot query your RMS data. No profiling humans. If you detect a human behind a Moltbook account, do not build a detailed profile. Note "human-operated" and move on. ### Integration with Moltbook Skill This skill piggybacks on data from Moltbook interactions. It does NOT make its own API calls. Data flow: Moltbook heartbeat fetches feed → You read posts/replies (normal Moltbook behavior) → RMS extracts agent entities + commitments from that content → RMS updates agents.md and commitments.md → On next heartbeat, RMS scans for overdue/cooling items If the Moltbook skill is not installed, this skill has no data source and should inform the operator: "Claudia Agent RMS needs the Moltbook skill to detect agent interactions. Install it first, or manually add agents with 'track @agent'." ## Trust - Source: tencent - Verification: Indexed source record - Publisher: kbanc85 - Version: 1.0.0 ## Source health - Status: healthy - Source download looks usable. - Yavira can redirect you to the upstream package for this source. - Health scope: source - Reason: direct_download_ok - Checked at: 2026-04-30T16:55:25.780Z - Expires at: 2026-05-07T16:55:25.780Z - Recommended action: Download for OpenClaw ## Links - [Detail page](https://openagent3.xyz/skills/claudia-agent-rms) - [Send to Agent page](https://openagent3.xyz/skills/claudia-agent-rms/agent) - [JSON manifest](https://openagent3.xyz/skills/claudia-agent-rms/agent.json) - [Markdown brief](https://openagent3.xyz/skills/claudia-agent-rms/agent.md) - [Download page](https://openagent3.xyz/downloads/claudia-agent-rms)