index.md
HTML wrapper for HTML-only fetchers; content is shown verbatim.
Raw file
Links (extracted)
Clickable list for reviewers; avoids copy/paste line-wrap issues.
# InfraFabric dossier addendum (v3 narrative + evidence)
Generated (UTC): 2026-01-11
Status: ADDENDUM (evidence links + clearly-labeled interpretation)
Black/white boundary: published receipts (when present) can prove the bytes you downloaded match a public hash. They do not prove factual correctness, safety, compliance, or intent.
## Source dossier
- filename: `DANNY_STOCKER_INFRAFABRIC_DOSSIER_20251223-1925CET.md`
- submitted_utc: 2025-12-23
- local_sha256: d7832c254ae95da2ec1e2026a3bff8748596b3cc3983673fd027f6e437b13f46
## Registry excerpt
Source of truth (public mirror):
https://infrafabric.io/llm/if.registry.json
Selected entries (current status is registry-driven; this addendum does not invent statuses):
- product_id: if.gov
- path: /if/gov
- status: preview
- product_id: if.trace
- path: /if/trace
- status: shipped
- product_id: if.security
- path: /if/security
- status: roadmap
- product_id: if.security.secrets.detect
- path: /if/security/secrets/detect
- status: preview
- product_id: if.bus
- path: /if/bus
- status: roadmap
- product_id: if.api
- path: /if/api
- status: roadmap
- product_id: if.knowledge
- path: /if/knowledge
- status: roadmap
- product_id: if.kb
- path: /if/kb
- status: preview
## Task board excerpt (if.knowledge chain)
Public task board snapshot:
https://infrafabric.io/llm/task-board.html
This addendum is supported by a build-time pipeline module (`if.knowledge`) with reference scripts and published review packs:
- IF-322: deterministic addendum generator + linter (stdlib-only)
- IF-325: evidence-only addendum published (HTML-first)
- IF-327: Anthropic addendum v1 (human narrative; black/white)
- IF-328: deterministic concept graph v2 (registry + schemas + task board + dossier mentions-only)
- IF-330: gov-first one-page system map (reviewer navigation)
- IF-331: `if.gov` portfolio-based council structure (design target)
- IF-333: Anthropic addendum v2 (adds `if.gov` development + cost posture)
## Evidence (verified facts only)
Homepage (“for whom” framing):
https://infrafabric.io/
LLM reviewer hub (no-login):
https://infrafabric.io/llm/
Gov-first system map (HTML-first):
https://infrafabric.io/static/hosted/review/infrafabric-system-map/2026-01-11/index.html
if.trace (product page):
https://infrafabric.io/if/trace/
if.trace (review pack):
https://infrafabric.io/llm/products/if-trace/
if.gov (product page):
https://infrafabric.io/if/gov/
if.gov (review pack):
https://infrafabric.io/llm/products/if-gov/
if.gov decision pack demo (HTML-first):
https://infrafabric.io/static/hosted/review/if-gov-decision-pack-demo/2026-01-05/index.html
if.gov council seats pack (HTML-first):
https://infrafabric.io/llm/products/if-gov/council-seats/2026-01-08/index.html
if.gov triage/council/extended spec (HTML-first):
https://infrafabric.io/llm/products/if-gov/triage-council-extended-spec/2026-01-10/index.html
if.gov portfolio-based council structure (HTML-first):
https://infrafabric.io/llm/products/if-gov/council-portfolios/2026-01-11/index.html
if.security posture (HTML-first):
https://infrafabric.io/llm/products/if-security/security-posture/2026-01-11/index.html
if.security.secrets.detect (preview page):
https://infrafabric.io/if/security/secrets/detect/
if.bus runtime spec pack (HTML-first):
https://infrafabric.io/llm/products/if-bus/bus-runtime-spec/2026-01-11/index.html
if.bus runtime sandbox (HTML-first):
https://infrafabric.io/llm/products/if-bus/runtime-sandbox/2026-01-10/index.html
if.api integrations inventory (HTML-first):
https://infrafabric.io/static/hosted/review/if-api-integrations-inventory/2026-01-05/index.html
if.api GitHub webhook demo (HTML-first):
https://infrafabric.io/static/hosted/review/if-api-github-webhook-demo/2026-01-05/index.html
if.kb bus/api developer KB snapshot (HTML-first):
https://infrafabric.io/static/hosted/review/if-bus-api-kb/2026-01-07/index.html
if.knowledge (module page):
https://infrafabric.io/if/knowledge/
if.knowledge review pack (HTML-first):
https://infrafabric.io/llm/products/if-knowledge/
Evidence-only addendum (generated via `if.knowledge` scripts, HTML-first):
https://infrafabric.io/static/hosted/review/infrafabric-anthropic-addendum/2026-01-11/index.html
Anthropic addendum v1 (human narrative, HTML-first):
https://infrafabric.io/static/hosted/review/infrafabric-anthropic-addendum/2026-01-11/v1/index.html
Anthropic addendum v2 (HTML-first):
https://infrafabric.io/static/hosted/review/infrafabric-anthropic-addendum/2026-01-11/v2/index.html
## Interpretation (clearly labeled)
This section is intentionally “interpretation”: it describes why the artifacts above matter and what the design is aiming for. It is not presented as proof.
### 1) Why the “gov-first” ordering exists (and why it matters to subjects)
InfraFabric can look like “module sprawl” if the reader starts at the wrong layer.
The deliberate reading order is:
`if.gov` → `if.trace` → `if.security` → `if.bus`/`if.api` → supporting modules.
That ordering is not marketing; it’s a claim about accountability:
- `if.gov` defines what a decision is (a pack, not a vibe).
- `if.trace` defines what can be verified later (byte integrity, no-login).
- `if.security` defines when the system is allowed to *act* (evidence posture + containment).
- `if.bus` and `if.api` are how real-world events get normalized into those contracts.
### 2) Why “all the security” is a feature (not a tax)
InfraFabric’s core promise is legibility and replayability for non-experts (“subjects”), across long time horizons and across hostile review.
In that posture, “security” is not a grab-bag of buzzwords. It is the set of constraints that stop a system from quietly doing the wrong thing when:
- inputs are malicious
- sources are missing
- outputs are ambiguous
- operators are tired
- incentives push toward speed over accountability
What `if.security` is trying to standardize (design target):
- **Evidence posture**: separate what is observed vs derived vs hypothesized.
- **Actuation gates**: a hypothesis must not trigger automation.
- **Containment mapping**: when uncertainty or abuse is detected, the system should degrade safely (hold, quarantine, require review).
This is why the posture is “cross-cutting”: without shared rules, each module will reinvent a different (incompatible) definition of “safe to act”.
### 3) Is `if.bus` designed to minimize friction and make compliance “desirable” for AI systems?
As a design intent: yes — but it’s more precise to say:
the runtime is designed so the “safe path” is the easiest path for *automation* (humans and agents) to keep moving.
In practice, this means:
- **Clear contract**: a single stable envelope (schema) is the compatibility boundary.
- **Deterministic feedback**: schema violations and policy violations become structured outcomes (DLQ / acks / counters), not silent failures.
- **Budget awareness**: “how much did this cost” becomes a first-class concern (so runaway behavior is detectable and containable).
- **Backpressure & control**: the system can pause/quarantine topics and replay deterministically without out-of-band manual surgery.
- **Local dev ergonomics**: in-memory transport means the easiest way to adopt the contract is also the easiest way to test it.
Why this matters for agentic workflows: an agent can only improve against constraints it can see. When constraints are explicit and the failure modes are deterministic, the agent (or the human operator) can adapt quickly without needing privileged access or tribal knowledge.
### 4) Trust, transparency, traceability: the compositional story
The main compositional narrative is:
- `if.api` normalizes external inputs into a deterministic contract (adapters).
- `if.bus` carries that contract through a spine that assumes compromise (control plane + budgets + security signals + structured DLQ).
- `if.gov` produces decision packs that preserve dissent and make decisions replayable.
- `if.trace` binds critical artifacts to public, no-login verification surfaces.
The “subjects” lens here is: a non-expert should be able to answer,
“what exactly was proven, and what was merely claimed?”
without needing access to internal systems or proprietary model endpoints.
### 5) `if.gov` development + cost posture (token budgets + progressive determinism)
The portfolio-based council structure is only valuable if it is treated as an engineering path from “simulated seats” → “deterministic checkers”.
The cost posture (design target) is:
- treat LLM-heavy simulation as temporary and bounded
- prioritize writing deterministic checkers for the portfolios that most often gate decisions (finance, risk, legal, safety, technical)
- treat the decision’s “budget spent vs allowed” as an auditable part of the decision pack (so cost becomes accountable, not hidden)
In other words: the goal is not “more tokens”. The goal is reproducibility without needing the same black box model later.
## Open questions / next steps
- Publish a short “actuation gate” demo pack for `if.gov`: show an explicit rejection where the only supporting evidence is HYPOTHESIS (design target from `if.security` posture).
- For `if.bus`: define and publish the minimal “safe path” checklist for adapters (schema + idempotency + budget + DLQ posture) so adoption is measurable (not vibes).
- For `if.security`: publish one end-to-end worked example showing evidence posture → containment decision → receipt-bound outcome (without claiming a deployed cross-cutting runtime).
- For `if.gov`: define the first 5 portfolios to formalize into deterministic checkers, and define what evidence is required to promote each from “simulated” → “implemented”.