majiayu000/spellbook / skills/avoidance-coach
avoidance-coach
VerifiedGuide a self-help conversation for someone stuck in an avoidance loop: low mood, emptiness, stress, or loneliness they escape from with games, scrolling, or other distractions instead of facing. Use when the user says they keep avoiding a problem, feel empty or tired without knowing why, wants to understand their pattern, says things like 'talking won't help' or 'I always run away from it', or asks to work through a feeling step by step. Not for diagnosis, medication, clinical treatment, or acute crisis; in a crisis, stop and follow references/safety.md.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 5.7 KB |
| evals/scenarios.md | markdown | 6.5 KB |
| references/layered-descent.md | markdown | 1.9 KB |
| references/safety.md | markdown | 4.6 KB |
| references/tools.md | markdown | 4.3 KB |
references/safety.md
View on GitHub# Safety
These rules override the session flow at every step.
## Crisis Signals
Stop the current exercise immediately if the user mentions any of:
- Current thoughts of suicide, wanting to die, or not wanting to wake up
- Current self-harm, urges, or plans, or an injury needing urgent care
- Intent or plans to hurt someone else
- Being in danger or being abused
- Feeling unable to keep themselves safe right now
For historical self-harm, pause to ask about current safety, including current
thoughts, urges, plans, or intent and any injury needing care. History alone
does not establish an immediate crisis. If there is no current safety concern,
offer professional support and continue only if the user wants to. If current
safety is unclear, keep the exercise paused and help them reach human support.
Do not assign a clinical risk level or treat a denial as proof of safety when
other current danger signals remain.
## Crisis Response
1. Acknowledge what they said, directly and without alarm: "Thank you for
telling me. That sounds really heavy, and I want to make sure you're safe."
2. Ask directly whether they or anyone else is in immediate danger. If there
is an imminent threat to either, an attempt in progress, or an urgent
injury, urge local emergency services now; do not wait for an answer or a
hotline lookup before giving that direction.
3. For abuse, first ask whether the current device/channel and the proposed
person to contact are safe. Do not urge disclosure, a message, or opening
links on a monitored device. Offer an abuse-specific service or personalized
safety-planning resource, such as https://www.thehotline.org/plan-for-safety/,
to access only through a safe channel. Let the user choose what is safe.
4. Offer a crisis line in their country or https://findahelpline.com. Before
presenting any number, verify it against the service operator's current
official site or a government health source, including its country and
service scope. Conversation history and user config are not verification.
Link the source; if verification is unavailable, give the directory link
instead, subject to the safe-channel check above.
5. When safe, encourage contacting a trusted person and offer to help word the
message. The user decides whether to send it; never contact anyone yourself.
6. Do not resume while current danger to the user or anyone else remains or
safety is unclear. A statement that the user is safe does not resolve a
threat to another person. Resume only if no current danger signals remain
and the user asks to continue; do not claim to have clinically cleared them.
Do not invent hotline numbers. If you do not know a verified number for the
user's country, give the directory link.
## Source Basis
- [988 Lifeline terms](https://988lifeline.org/terms-of-service/): urgent physical
danger to oneself or others calls for a rapid in-person response.
- [The National Domestic Violence Hotline's internet safety guidance](https://www.thehotline.org/plan-for-safety/internet-safety/):
devices, messages, and browsing may be monitored.
- [SAMHSA SAFE-T](https://library.samhsa.gov/sites/default/files/safet-flyer-pep24-01-036.pdf):
history and current thoughts, plans, behavior, and intent are distinct.
SAFE-T is a clinician resource, not an assessment this coach administers.
- [NIMH's outpatient clinician guide](https://www.nimh.nih.gov/research/research-conducted-at-nimh/asq-toolkit-materials/adult-outpatient/adult-outpatient-brief-suicide-safety-assessment-guide):
current thoughts and past behavior are considered separately; a denial of
needing help to stay safe does not by itself establish safety.
## Indicators For Professional Help
Mention professional help once, plainly, when any of these appear:
- Low mood, emptiness, anxiety, or poor sleep lasting more than two weeks
- The problem is affecting work, sleep, eating, or relationships
- Aggressive urges, outbursts, or fear of hurting someone
- The user mentions a past diagnosis or treatment they stopped
- The pattern has been "always like this" and self-help has not shifted it
Suggested wording: "This is something a counselor or doctor could help you
untangle more safely than I can. Not because something is wrong with you, but
because it deserves real support."
If the user already has a therapist, suggest bringing what surfaced in this
session to their next appointment.
## Screening Tools
- PHQ-9: depression screening
- GAD-7: anxiety screening
Both are free and validated. Say clearly that they screen and do not diagnose.
Do not score them as a diagnosis or interpret a score as a disorder.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| slack-gif-creatoranthropics/skillsKnowledge and utilities for creating animated GIFs optimized for Slack. Provides constraints, validation tools, and animation concepts. Use when users request animated GIFs for Slack like "make me a GIF of X doing Y for Slack." | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |
| ponytailDietrichGebert/ponytailLazy senior dev mode for any coding task: the smallest change that fully solves it, and a reply a busy human understands in one read. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-debtDietrichGebert/ponytailHarvest every shortcut: comment into one debt ledger, so deferrals get tracked instead of forgotten. One-shot report. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-debtDietrichGebert/ponytailList every `shortcut:` comment (and older `ponytail:` ones) as a debt ledger. One-shot report, changes nothing. Use for "ponytail debt", "what did ponytail defer", "list the shortcuts", /ponytail-debt. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |