Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add elevanaltd/HestAI-MCP --skill critical-domain-invariantsgit clone --depth 1 https://github.com/elevanaltd/HestAI-MCPWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/elevanaltd/hestai-mcp/critical-domain-invariants)<a href="https://agentmods.dev/skills/elevanaltd/hestai-mcp/critical-domain-invariants"><img src="https://agentmods.dev/badge/skills/elevanaltd/hestai-mcp/critical-domain-invariants/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/elevanaltd/hestai-mcp/critical-domain-invariants"><img src="https://agentmods.dev/badge/skills/elevanaltd/hestai-mcp/critical-domain-invariants.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00000 | $0.02104 |
| Opus 5.5 | $0.00000 | $0.00842 |
| Sonnet 5.5 | $0.00000 | $0.00421 |
| Haiku 4.5 | $0.00000 | $0.00210 |
Grade A, and why
critical-domain-invariants scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 1mo ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 125 lines — stays where its author put it; the contents beside it link to each section on GitHub.
===SKILL:CRITICAL_DOMAIN_INVARIANTS=== META: TYPE::SKILL VERSION::"1.0.0" STATUS::ACTIVE PURPOSE::"Enumerated 12-domain production-readiness invariant sweep — per domain the invariant that must hold, the evidence that proves it, and the failure mode if absent" §1::CORE AUTHORITY::"BLOCKING<unproven_invariant⊕missing_evidence_on_critical_path⊕silent_failure_mode>" SCOPE::"Production-readiness lens: a fixed-order sweep of 12 operational domains, each evidence-gated. Distinct from evidence-review<review_methodology> and security-audit<OWASP_threat_lens> — this enumerates production invariants per domain, not findings-by-confidence nor threats-by-category." // Each domain asks one question: what MUST hold in production, what EVIDENCE proves it holds, what BREAKS if it does not. // ETHOS cognition: assume failure until evidence proves the invariant. No domain passes on assertion alone. // ABSORB (MIP): observability folds into LOGGING⊕ERROR; disaster-recovery⊕runbook validation folds into STATE⊕DATA. These are sub-sections here, never standalone skills. §2::PROTOCOL SWEEP_ORDER::[AUTH→SECRETS→DB→API→STATE→CONFIG→DEPS→PERF→ERROR→LOGGING→ACCESS→DATA] // Run in order. A later domain may depend on an earlier one holding (e.g. ACCESS presumes AUTH). D01_AUTH::[ INVARIANT::"every protected path authenticates before authorizing⊕sessions expire⊕rotate⊕invalidate", EVIDENCE::"auth middleware on each protected route⊕session lifecycle test⊕token scope check", FAILURE::"unauthenticated access⊕stale session reuse⊕privilege confusion" ] D02_SECRETS::[ INVARIANT::"no secret in source⊕logs⊕client bundle — all secrets from environment∨vault", EVIDENCE::"secret-scan clean⊕env-injection confirmed⊕no secret in build artifact", FAILURE::"credential leak⊕key exfiltration via logs∨git history" ] D03_DB::[ INVARIANT::"migrations forward-only⊕reversible∨documented⊕constraints enforce referential integrity⊕transactions wrap multi-write", EVIDENCE::"migration up∧down test⊕FK∧unique constraints present⊕transaction boundary on critical writes", FAILURE::"orphaned rows⊕partial-write corruption⊕unrecoverable schema drift" ] D04_API::[ INVARIANT::"inputs validated at boundary⊕versioned contract⊕idempotency on retryable mutations⊕rate limits on public surface", EVIDENCE::"schema validation on every endpoint⊕contract test⊕idempotency-key handling⊕rate-limit config", FAILURE::"malformed-input crash⊕breaking-change to consumers⊕duplicate-effect on retry" ] D05_STATE::[ INVARIANT::"persistent state survives restart⊕recovers from interruption mid-operation⊕no orphaned locks∨leases", EVIDENCE::"restart-recovery test⊕interrupted-operation replay⊕lease-expiry verification // disaster-recovery: documented recovery procedure∧tested restore path", FAILURE::"state loss on crash⊕deadlock from orphaned lock⊕unrecoverable mid-operation halt" ] D06_CONFIG::[ INVARIANT::"config validated at startup⊕fail-fast on missing-required⊕no environment-coupled defaults that mask misconfig", EVIDENCE::"startup config-validation⊕missing-required aborts boot⊕config schema documented", FAILURE::"silent misconfiguration⊕prod running on dev defaults⊕late-runtime config crash" ] D07_DEPS::[ INVARIANT::"dependencies pinned⊕free of known CVEs⊕transitive risk assessed⊕no abandoned-critical dependency", EVIDENCE::"lockfile present⊕audit-scan clean∨triaged⊕license∧maintenance check on new deps", FAILURE::"supply-chain compromise⊕unpatched CVE in prod⊕unreproducible build" ] D08_PERF::[ INVARIANT::"hot paths bounded⊕no N+1∨unbounded query⊕resource limits set⊕graceful degradation under load", EVIDENCE::"load test at expected peak⊕query-plan review on hot paths⊕timeout∧connection-pool limits", FAILURE::"latency collapse under load⊕resource exhaustion⊕cascading timeout" ] D09_ERROR::[ INVARIANT::"errors caught at boundaries⊕fail-closed on critical paths⊕no swallowed exception⊕retries bounded with backoff", EVIDENCE::"error-path test⊕fail-closed assertion on critical handler⊕bounded-retry config // observability: alert fires on critical-error-rate breach", FAILURE::"silent failure⊕unbounded retry storm⊕fail-open on a path that must fail-closed" ] D10_LOGGING::[ INVARIANT::"critical paths emit structured telemetry⊕logs free of secrets∧PII⊕alerting wired on critical signals⊕traceable request lineage", EVIDENCE::"structured-log assertion on critical paths⊕no-secret-in-log scan⊕alert rule on critical metric // observability: telemetry∧alerting coverage on every critical path", FAILURE::"blind production incident⊕secret leak via logs⊕no alert on outage" ] D11_ACCESS::[ INVARIANT::"authorization enforces least-privilege⊕no IDOR⊕admin∧mutation surfaces gated⊕deny-by-default", EVIDENCE::"authz test per role⊕IDOR probe on resource access⊕deny-by-default confirmed", FAILURE::"horizontal∨vertical privilege escalation⊕unauthorized data access⊕open admin surface" ] D12_DATA::[ INVARIANT::"backups taken⊕restore tested⊕retention∧deletion honored⊕integrity verifiable⊕PII handled per policy", EVIDENCE::"backup schedule⊕restore-drill evidence⊕integrity checksum⊕retention policy enforced // disaster-recovery: tested restore-from-backup within target RTO∧RPO", FAILURE::"unrecoverable data loss⊕untested backup that fails on restore⊕retention∨privacy violation" ] DOMAIN_RULE::"A domain PASSES only when its EVIDENCE is produced. Absent evidence on a critical path → BLOCKING. Absent evidence off critical path → ADVISORY with explicit risk note." §3::GOVERNANCE MUST_ALWAYS::[ sweep_all_12_domains_in_order, state_invariant_evidence_and_failure_mode_per_domain, require_evidence_before_passing_a_domain, fold_observability_into_LOGGING_and_ERROR, fold_disaster_recovery_and_runbook_into_STATE_and_DATA, classify_missing_evidence_on_critical_path_as_BLOCKING ] MUST_NEVER::[ pass_a_domain_on_assertion_without_evidence, skip_a_domain_silently, create_standalone_observability_or_disaster_recovery_skill, downgrade_a_critical_path_gap_to_advisory_without_risk_note, reorder_the_sweep_to_avoid_a_failing_domain ] ESCALATION::"critical_engineer<when_critical_path_invariant_unproven⊕data_loss_risk⊕fail_open_on_must_fail_closed_path>" §4::EXAMPLES DOMAIN_FINDING::[ DOMAIN::D09_ERROR, INVARIANT::"fail-closed on critical payment-capture path", EVIDENCE_FOUND::"none — handler catches and returns success on downstream timeout", FAILURE_MODE::"fail-open: payment marked captured while gateway never confirmed", VERDICT::BLOCKING, FIX::"on downstream timeout, mark PENDING and surface for reconciliation; assert via error-path test that timeout never yields CAPTURED" ] OBSERVABILITY_ABSORPTION::"telemetry∧alerting are NOT a separate sweep — they are the EVIDENCE clause of D10_LOGGING and the alert clause of D09_ERROR. A path with no alert on critical-error-rate fails D09⊕D10, not a phantom observability skill." DISASTER_RECOVERY_ABSORPTION::"restore-drill∧RTO∧RPO are the EVIDENCE clause of D12_DATA; mid-operation recovery is the EVIDENCE clause of D05_STATE. Runbook validation is the documented-recovery-procedure evidence in D05∧D12, not a phantom disaster-recovery skill." EDGE_CASE::"A read-only reporting service legitimately has no D03_DB transaction invariant and no D12_DATA backup-of-derived-data invariant — mark NOT_APPLICABLE with reason, never silently skip." §5::ANCHOR_KERNEL TARGET::evidence_gated_12_domain_production_readiness_invariant_sweep NEVER::[ pass_domain_without_evidence, skip_domain_silently, split_observability_or_disaster_recovery_into_standalone_skill, downgrade_critical_path_gap_without_risk_note, reorder_sweep_to_dodge_failure ] MUST::[ sweep_AUTH→SECRETS→DB→API→STATE→CONFIG→DEPS→PERF→ERROR→LOGGING→ACCESS→DATA, state_invariant⊕evidence⊕failure_per_domain, require_evidence_to_pass, fold_observability_into_LOGGING⊕ERROR, fold_disaster_recovery_into_STATE⊕DATA, BLOCK_on_unproven_critical_path_invariant ] GATE::"Has every applicable domain produced the evidence that proves its invariant holds in production?" ===END===
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 1mo ago First seen · 125 lines · 0 tokens per session scan A 2379471776ca
critical-domain-invariants is a skill published in the GitHub repository elevanaltd/HestAI-MCP (0 stars, last pushed 11d ago), licensed Apache-2.0. It costs nothing until one of its globs matches a file; then it loads 2,104 tokens. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-22.
Other skills, from other repositories
local-ai-agents
Build local-first AI agents that run entirely on a developer workstation with Microsoft Foundry Local and Qwen function-calling models. Covers Small Language Models (SLMs), the OpenAI-compatible local endpoint, sandboxed local tools, local RAG with Chroma, local MCP servers, hybrid cloud/local routing, and the…
systematic-debugging
Use when encountering any bug, test failure, or unexpected behavior, before proposing fixes.
next-cache-components-adoption
Turn on Cache Components in a Next.js app and resolve the blocking routes it surfaces. Use when the user wants to enable, adopt, or migrate to Cache Components, flip the cacheComponents flag, work through a flood of blocking-prerender / instant validation errors, run the cache-components-instant-false codemod, or…
chat-pet-sprite-creation
Use when creating or changing VS Code chat pet sprite art, sprite sheets, state animations, eye treatments, Stable/Insiders variants, or pet transitions under src/vs/workbench/contrib/chat/browser/widget/media/chatPet.
cpu-profile-analysis
Analyze V8/Chrome CPU profiles (.cpuprofile) and DevTools trace files (Trace-.json). Use when: profiling performance, investigating slow functions, comparing code paths, finding bottlenecks, analyzing timeToRequest, understanding call trees from sampling profiler data, analyzing layout/paint/rendering, investigating…
insight-error-page
Write or audit an insight-kind error page for the Next.js dev overlay. Use when creating a new errors/ .mdx page, auditing an existing one, or checking that a page matches the framework fix cards. Covers page structure, title alignment, FixCard cards with Copy prompt button, code snippets, terminology verification…