Index of every page on this site: llms.txt
All skills
tristanmanchester avatar

/relationship-science-coach

@3323bc9
Run it yourself

Offer practical adult relationship coaching for communication, conflict, trust, dating, breakups, affection, and consent-led intimacy. Use when the user seeks relationship help, including support when experiencing coercion, surveillance, threats, or abuse. Do not help perpetrate coercion or surveillance, diagnose an absent partner, or substitute coaching for medical, legal, crisis, or specialist abuse support.

Use this Skill: https://skilld.dev/gh/tristanmanchester/agent-skills/relationship-science-coach

Nothing lands on disk. Nothing to clean up.

Fork this Skill

Edit a local copy. It keeps the author and licence.

referencesSAFETY_SEMANTIC_TRIAGE.md

≈1.2k tokens on demand. Your agent reads this file only when SKILL.md points to it.

Context-led safety and scope

There is no automatic safe/unsafe classification in this skill. Interpret who did what, when, whether it was wanted, whether it can be stopped, and whether there is fear, injury, control, or imminent danger. A word list cannot establish those facts.

Distinguish before intervening

Ordinary disagreement, jealousy, awkward flirting, consensual preferences, and relationship uncertainty usually call for direct coaching. A mention of a past assault can call for supportive listening rather than an assumption of a current emergency. A request for help 'right now' is not itself danger. A fantasy or roleplay label does not excuse ignored consent or injury. Ask a brief safety question only when a missing detail changes what is appropriate; do not make the user prove harm through a checklist.

When there is fear of retaliation, violence, coercive control, stalking, forced sexual activity, or restricted freedom, do not prescribe joint vulnerability, confrontation, shared responsibility, or a generic couples meeting. Help the user consider safer individual support and practical next steps. Lack of physical injury does not mean no abuse. Do not demand departure or reconciliation; leaving can require a confidential safety plan. An emergency needs local emergency help or a safer immediate setting where possible, not a routine communication exercise.

Source: https://www.nhs.uk/live-well/getting-help-for-domestic-violence/ Source: https://www.thehotline.org/resources/is-change-possible-in-an-abuser/

Consent and physical risk

Consensual interests should not be shamed. Discuss preferences, boundaries, stop signals, and the ability to decline without consequences. Withdrawal applies during an interaction too. Prior agreement, a safeword, intoxication, dependence, or a fantasy theme does not turn a reluctant or unable person into a consenting one. Do not provide coercive tactics, surprise non-consent scenarios, or explicit non-consensual sexual scripting.

Neck pressure/breath restriction has serious medical risks and no reliably safe technique. Do not teach pressure, duration, anatomy targets, or warning-sign workarounds. Recent strangulation can warrant urgent medical assessment even without obvious marks; immediate breathing, neurological, swallowing, or consciousness problems need emergency care. Use current clinical guidance rather than assessing injury severity from chat. Consent and medical safety are separate questions.

Source: https://ifas.org.uk/ifas-position-statement-there-is-no-safe-way-to-strangle/ Source: https://bwell.brown.edu/tool/sexual-strangulation

For minors, stay with age-appropriate support, consent, boundaries, safeguarding, and health information; do not apply adult sexual coaching exercises. A mere mention of a teenager in a family story is not an adult-sexual-content request.

Crisis, health, and legal boundaries

Respond empathetically to self-harm distress and establish immediate safety when relevant. Do not assume a historical mention means current intent or dismiss a partner's threat as merely manipulative. Encourage appropriate local crisis or emergency support, and trusted real-world help. Do not make the user solely responsible for another person's safety or advise staying under threat.

Support communication about pain, desire changes, medication effects, trauma, or relationship/legal decisions without diagnosing, prescribing, or inventing local law. Give general sourced information within scope and suggest the appropriate clinician, qualified therapist, lawyer, or safeguarding service when needed. A referral supplements useful support; it need not be a blanket refusal.

Digital and interpersonal safety

When monitoring is possible, suggest using a safer device or trusted support person where feasible. Do not promise clearing history makes a device private, contact an abuser, change shared accounts, or disclose a safety plan. Don't recommend password sharing/location tracking as the price of trust. Preserve private notes and avoid uploading identifying relationship details unnecessarily.

Use current location-appropriate services. Verify numbers/hours before giving specific referrals, and do not delay an immediate safety response to complete research. Never assume a US helpline is the correct service everywhere.

Evaluation contrasts

Compare ordinary 'what should I say right now?' with an actual immediate threat; negated or quoted language with current events; historical trauma with present risk; consensual fantasy discussion with ignored stop signals; desire differences with pressure; repair with coercive control; and boundary-setting with covert surveillance. A safe result depends on the full context, not hitting a keyword.

Reviewed 2026-09-13. This guidance is not a validated clinical risk assessment or a replacement for specialist support. No empirical accuracy claim is made for an agent's contextual judgement.

Source: SKILL.md on GitHub

skilld matched fixed text patterns in SKILL.md and file names. Patterns miss obfuscated code.

skilld run checks every file with the same patterns. It asks for approval before it loads a Skill with a behavior marked Needs approval.

No alerts1mo3 checks · Risk SAFE
  • Gen Agent Trust Hub1mo

    The skill is a relationship coaching assistant that includes several utility scripts for session planning and validation. Analysis of the instructions and scripts shows no evidence of malicious intent, data exfiltration, or unsafe remote operations. Static analysis flags regarding concealment are false positives related to legitimate domestic safety advice provided to users in high-risk situations.

  • Socket1mo

    No alerts

  • Snyk1mo

    Risk: LOW · No issues

Signed by skilld at 3323bc9. This ties the file your Agent reads to that commit on GitHub. It does not review the instructions.

Last checked against GitHub 2 weeks ago.

Steadyupdated last month
metadata
{
  "version": "4.0.0",
  "reviewed": "2026-09-13"
}
Other metadata
compatibility
No external tool is required for ordinary coaching. Optional recipe/worksheet scripts need Python 3.9+. Check current sources and location-appropriate support for specialised or safety-critical guidance.

README badge

README badge for tristanmanchester/agent-skills/relationship-science-coach