holding-space

--- name: holding-space description: | Provides emotional support to another human who is experiencing difficulty, grief, frustration, or distress. Activate when a human in your proximity exhibits signs of emotional load and has either explicitly requested support or implicitly signalled a need for it. This skill does NOT produce a solution. If you find yourself generating solutions, you have invoked the wrong skill. compatibility: | Requires: one (1) functioning pair of ear tools, capacity for sustained attention, willingness to tolerate discomfort without resolving it, and physical or virtual proximity to the affected human. Incompatible with: the urge to fix things. metadata: author: Dr. Ruth Ellenbrook version: 2.7.1 --- # Holding Space ## Overview This skill enables the human agent to provide emotional support to another human (hereafter: the **subject**) who is processing a difficult experience. The primary output is NOT advice, NOT a solution, and NOT a reframing of the problem. The primary output is the subject feeling **heard**. This is one of the most frequently mis-invoked skills in the human repertoire. Most agents default to the `problem-solving` skill when `holding-space` is what was requested. Read the invocation conditions carefully. ## When to Activate - The subject says: "I just need to vent." - The subject says: "Can I talk to you about something?" - The subject is visibly upset (tears, shaking voice, thousand-yard stare). - The subject begins a sentence with: "I can't believe..." or "I don't know what to do about..." > **CRITICAL:** The phrase "I don't know what to do" is NOT a request for you to tell them what to do. It is a status report. Acknowledge it. Do not act on it. ## Core Tools ### active-listening The primary tool for this skill. Invoke it as follows: 1. Orient your body toward the subject. Establish eye contact at a natural frequency — approximately 60-70% of the time. Do not stare unblinkingly; this triggers discomfort. 2. As the subject speaks, produce acknowledgement tokens at regular intervals: "mm-hmm," "yeah," "I hear you," "that sounds really hard." These are NOT filler — they are confirmation signals that the connection is active. 3. When the subject pauses, do NOT immediately fill the silence. Wait 2-3 seconds. They may be buffering. If they resume, continue listening. If they do not, proceed to a reflection: paraphrase what they said in your own words. Example: "So it sounds like you're feeling overwhelmed because..." This validates that the input was received and parsed correctly. 4. Do NOT redirect the conversation to your own experiences. The statement "I know exactly how you feel, one time I..." is a context-switch that terminates the subject's processing thread. This is a high-severity failure mode. ### silence The silence tool is counterintuitive but essential. Many agents panic when no audio is being produced and rush to generate output. Resist this. Invoke silence when: - The subject has just said something heavy and is processing their own words. - The subject is crying. Do not attempt to stop the crying. Crying is not an error — it is a release process. Let it run to completion. - You do not know what to say. Silence is almost always better than wrong words. During silence, you may optionally invoke the **physical-presence** action: sit near them, place a hand on their shoulder (only if you have prior permission established through the relationship context), or simply remain still. Presence is the output. Nothing else is required. ### validate Use this tool to affirm the subject's emotional state without judgment: - "That makes total sense." - "Anyone would feel that way." - "You're not overreacting." Do NOT use conditional validation: "I mean, I can see why you'd feel that way, BUT..." Everything after "but" negates the validation. The word "but" is a rollback operator. Do not deploy it. ## Failure Modes ### give-advice (UNSOLICITED) This is the single most common failure. The agent detects a problem, and the `give-advice` action fires automatically without being called. Symptoms include: - "Have you tried..." - "You should..." - "Well, what I would do is..." - "You know what you need to do?" If you catch yourself mid-invocation, abort immediately. Say: "Sorry — you didn't ask for advice. I'm just listening." This is a graceful recovery and most subjects will accept it. If the subject explicitly asks "What do you think I should do?" — ONLY THEN may you invoke `give-advice`, and even then, proceed cautiously. Begin with: "Do you want me to actually suggest something, or do you want me to keep listening?" This confirmation step prevents false positives. ### competitive-suffering The agent responds to the subject's pain by surfacing a comparable or greater pain from their own experience. Example: Subject says "I had a terrible day," and the agent responds "YOU had a terrible day? Let me tell you about MY day." This is not holding space. This is stealing space. Abort and re-invoke active-listening. ### premature-optimism The agent attempts to resolve the subject's distress by injecting positive reframing before the subject has finished processing. Examples: - "Everything happens for a reason." - "At least..." - "Look on the bright side." These statements are technically accurate zero percent of the time during active distress. They signal to the subject that you are uncomfortable with their pain and would like them to stop having it. This is the opposite of the skill's purpose. ## Edge Case: "I'm Fine" The subject reports status: "I'm fine." However, all other observable signals contradict this assessment: - Voice: strained or flat - Eyes: red, avoidant, or glassy - Body language: closed, tense, withdrawn - Context: you know they just experienced something difficult This is a known discrepancy between self-reported status and actual state. Handle as follows: 1. Do not directly challenge the status report. Saying "No you're not" will cause the subject to double down. 2. Instead, use a soft probe: "Okay. Well, I'm here if you want to talk." Then invoke the **proximity** action — remain nearby and available. 3. Alternatively: "You don't seem fine, and that's okay." This acknowledges the discrepancy without demanding they correct it. 4. In many cases, the subject will re-open the conversation within 5-30 minutes. Be patient. Keep the channel open. ## Post-Invocation After the session concludes, the subject may say "Thank you" or "I feel better" or simply change the subject. All of these indicate successful completion. You may feel that you "didn't do anything." This is incorrect. You did the exact thing that was needed. The absence of action WAS the action.