SuperAPI AI Directory

health-check

Deployment & operations

View repository

Updated 10-16 05:29

About this skill

health-check

Self-Evolving Skill: This skill improves through use. If instructions are wrong, parameters drifted, or a workaround was needed — fix this file immediately, don't defer. Only update for real, reproducible issues.

Self-test + quota for the po plugin, via the TS core.

env -u HTTPS_PROXY -u HTTP_PROXY bun "$(cc-plugin-root pushover-commander)/skills/_lib/pushover_core.ts" doctor   # full self-test (JSON)
env -u HTTPS_PROXY -u HTTP_PROXY bun "$(cc-plugin-root pushover-commander)/skills/_lib/pushover_core.ts" quota     # {limit, remaining, reset}

doctor checks: creds (op→Keychain), /users/validate.json, monthly quota, the expected custom sounds (po_fanfare/po_uplift/po_celebrate), deps (bun/uv/Chrome), and the audit log size. quota prints remaining messages and warns when below quota_warn_remaining (see pushover_api_limits.json).

Reliability primitives (apply to every send)

  • Limits SSoT — skills/_lib/pushover_api_limits.json is the single source for all Pushover limits/rules (message 1024, title 250, url 512, attachment 5 MB, sound <500 KB/≤30 s, app name 20, desc 500, html⊕monospace, priority-2 needs retry+expire). Edit there, nowhere else.
  • Preflight — pushover_core.ts validates every send against pushover_api_limits.json and turns Pushover's silent failures into explicit output: refuses (url>512, attachment>5 MB, html+monospace) — override with --force — and warns (message/title truncation, sound not in account).
  • Audit trail — every send appends a UUID-keyed JSONL line to ~/.local/state/pushover/po-audit.jsonl (ts, uuid, app, priority, status, request, receipt, remaining, message_len, errors). Retrieve with grep <uuid> ~/.local/state/pushover/po-audit.jsonl | jq ..
  • Quota guard — each send reads X-Limit-App-Remaining and warns when low.
  • Retry/backoff — transient network / 5xx are retried (×3, backoff) before failing.

Post-Execution Reflection

After this skill completes, check before closing:

  1. Did the check surface real failures (credentials, quota, dependencies) with no false green? If a broken subsystem reported healthy, fix the check.
  2. Was remaining quota reported accurately? A stale or wrong quota reading needs fixing before it misleads.

Only update if the issue is real and reproducible — not speculative.