posthog-observability
Audit already-collected PostHog evidence for production instrumentation, errors, releases, performance, conversions, privacy, and bounded draft-fix eligibility without contacting providers.
- License 许可
- MIT
- 最近更新
- 2026-11-14
- Stars
- ★ 0
- Forks
- 0
01 前言 Front Matter
- name
- posthog-observability
- description
- Audit already-collected PostHog evidence for production instrumentation, errors, releases, performance, conversions, privacy, and bounded draft-fix eligibility without contacting providers.
02 资源文件 1
- 参考文档agents/openai.yaml272 B
03 SKILL.md 正文
PostHog Observability
Use this skill to turn repository-specific PostHog configuration and exported evidence into a deterministic operational report. It is a read-only evaluator: do not query live PostHog, mutate a project, create alerts, open a draft fix, or write to any external system.
Require a production contract
Pass explicit repository policy and collected evidence to
auditPostHogObservability in src/posthog-observability.mjs. Verify:
- production-only capture and exact canonical hosts;
- anonymous-to-authenticated identity continuity;
- named conversion events and production exception capture;
- immutable release identity and a source map uploaded for that release;
- required Web Vitals;
- session replay with text masking, media blocking, and no sensitive event properties.
Missing evidence stays unproven or becomes a finding. Never infer broken instrumentation from low volume alone.
Classify before alerting
Classify observations as eligible production signals, production errors, preview contamination, noncanonical traffic, release mismatch, bot traffic, or expected validation. An insufficient eligible sample creates an investigation; it does not silently become an instrumentation failure.
An alert is actionable only when its route names an owner, private destination,
runbook, and threshold. Keep raw event payloads in the provider. Shared and
Check-in reports must never include replay URLs, PII, event properties, or
private payloads; consume only the returned sanitized checkInFindings.
Bound automation
Prepare a draft-fix intent only when the evidence packet contains all of:
- deterministic reproduction at an exact Git revision with concrete steps;
- a bounded root cause naming its module and explanation;
- passing real verification (the repository's verification CLI, the browser or computer use) with its evidence path or URL.
The audit decision is prepare-draft-fix, but the audit never creates the
draft. Any missing or ambiguous element produces investigate.
Do not remove Sentry or another provider until the same repository proves every declared parity capability. Even proven parity returns a separate retirement decision; it never removes the provider automatically.
Preserve readOnly: true, empty externalWriteIntents, and empty
externalSideEffects in every result.
04 同仓库更多 Skills
working-backwards
Define a software feature through PRFAQ and staged documents when the user requests working backwards before coding.
flow-implement
Implement a requested software change through verification, corrections, and its authorized delivery endpoint.
development-steward
Run the scheduled, read-only weekly maintenance review for the six explicitly opted-in repositories and prepare one concise private report for Check-in.
grill-with-docs
Run one recoverable interview for functional, visual, or mixed software grills; reuse equivalent answers and offer the shared HTML questionnaire.
global-agent-guardrails
Install, audit, test, or update the Development System's global destructive-command guard for Codex, Claude Code and T3 Code. Use when configuring agent safety hooks or investigating whether dangerous shell operations are blocked.
impeccable
Design or refine frontend UI using project context and a focused design playbook; also handles visual critique and polish.