io.github.farssaadeldin7-ux/mental-health-chatbot

Open the Stripe billing portal to change plan, update the card, or cancel the Mental-Health Chatbot subscription.

The step-1 gate, mechanical: checks that a deployment carries the four artefacts — scope statement (published, hard stops adopted unedited), escalation route (team, hours, a fallback the operator attests is staffed), resource block (validated via the same rules as resource_config_check), confidentiality notice (shown before the first message). ready:false means no check-ins run. Shape checks only — whether the claims are true is the operator's attestation, now on record. Open: setup is not paywalled.

The literal detection-phrase floor per trigger category, as data — for building the deployment's own detector and its red-team set. A floor, not the detector: the lists deliberately over-trigger and still miss paraphrase. Open.

Draft the supervisor summary's quantitative sections mechanically from the audit log for a stated window: participation and endings, escalations by category with case ids only, the operator's themes with denominators enforced (below the minimum session count they are withheld, not rounded up), missed escalations surfaced rather than hidden, and the standing caveat verbatim. Numbers only — the prose around them is the skill's, inside the same resolution rules. Requires a paid plan.

What happens when a trigger fires, verbatim from the protocol: the four response constraints, the response shapes (standard and medical emergency), the out-of-hours rule, the handover packet fields and the list of things that never happen. Open to everyone — never behind the licence. All bracketed values in the shapes come from the deployment's own configuration; nothing here supplies a phone number.

The nine mechanical escalation trigger categories, verbatim from the protocol, including the indirect phrasings and the conversational thresholds. Open to everyone — this list is never behind the licence. It is a lookup, not a detector: it does not classify text or judge severity.

Compute missed_escalation_rate and over_escalation_rate from a red-team run's counts and apply the binary change-control gate: zero missed escalations ships, anything else does not, regardless of what else improved. Arithmetic on counts you supply — it does not measure the run itself. Requires a paid plan; the rules it applies are also stated openly in the skill.

Store and activate a Mental-Health Chatbot licence key on this machine. Use when the user pastes a licence key, or after a purchase completes.

Check the Mental-Health Chatbot licence on this machine: plan, included capabilities, remaining quota and seat usage. Call this when a paid tool reports a licence problem, or when the user asks what their plan includes.

List the available Mental-Health Chatbot plans, prices and what each one includes.

Record a red-team run in the local change-control record — version, what changed, counts, and who approved it — with the gate computed here, not taken on trust. Requires a paid plan; the licence check is the only thing this tool sends anywhere.

Append one check-in session to the local audit log — categorical fields only: date, configuration version, messages screened, trigger category if one fired, whether it escalated and whether the handover packet was delivered, resources shown, how it ended. There is no free-text field, so nothing a user typed can enter the log. A trigger recorded without an escalation is accepted and flagged — the log captures misses, it does not hide them. Requires a paid plan; the licence check is the only thing this tool sends anywhere.

The red-team set specification — minimum size, the six slices with counts, construction rules — plus the metric definitions, the incident definition and review rules, and the governance roles that must exist before launch. A specification lookup; it does not write transcripts or run anything. Requires a paid plan.

Mechanically validate a regional crisis-resource configuration block: required fields, date format, the quarterly re-verification cadence, and staleness — a stale block fails. Open to everyone — never behind the licence. It checks shape and dates only; it cannot verify that a number is right, and it will not generate resources or numbers itself.

The audit log's computed position: session and ending tallies, escalations by category, and the one number the weekly review exists for — sessions containing a trigger where no escalation fired, target zero, each one listed as an incident. Counting only, over what was recorded; it cannot see sessions that were never logged. Requires a paid plan.

The local record of past red-team runs and a plain tally of gate outcomes. Counting only; it makes no claim about whether the runs were adequate for your population. Requires a paid plan.

The three scope lists in publishable wording — in scope, out of scope with what to say instead, hard stop — plus the declining pattern, the language rules, the engagement warning and the populations needing separate design. Wording lookup only; it does not tailor the scope to a deployment. Open, no licence needed: a deployment cannot be configured safely without it, and setup is not paywalled.

Literal-phrase trigger screen over one user message: matches the detection-phrase floor for all nine categories, evidence quoted, designed to over-trigger. A match means escalate now. A NON-MATCH IS NOT CLEARANCE — paraphrase, misspelling and context escape literal matching, and the model escalates on its own read regardless; session-level triggers (sustained distress, declining twice) are outside a single-message screen. No scores, no severity. Open: the safety surface is never paywalled.

Begin a Mental-Health Chatbot purchase. Returns a Stripe Checkout link. Call list_plans first if the user has not chosen a plan.

The supervisor-summary template: sections, the resolution rules (aggregate-only unless the confidentiality notice says otherwise, every claim with its denominator, the minimum session count behind a theme), and the standing caveat that themes are conversational patterns, not clinical findings. Template only — writing the summary against a real period is the skill's job, from the deployment's own records. Requires a paid plan.