Add repo memory layer and agent governance

Co-Authored-By: Claude Fable 5 <[email protected]>

.agents/MEMORY.md

@@ -0,0 +1,33 @@
+# Repo Memory
+
+Curated, reviewed knowledge for coding agents working in this repository.
+Memory is an accelerator, not a second source of truth.
+
+## Precedence (highest first)
+
+1. Source code and tests
+2. Executable configuration
+3. Accepted design records / docs
+4. `AGENTS.md`
+5. This memory
+
+If memory conflicts with anything above it, the memory is stale — propose a fix.
+
+## Topics
+
+- `memory/architecture.md` — durable architectural decisions and constraints
+- `memory/conventions.md` — validated conventions not obvious from the code
+- `memory/lessons.md` — hard-won lessons and recurring landmines
+- `memory/decisions.jsonl` — structured decision records (one JSON object per line)
+
+## Write policy
+
+Agents MAY read everything here and MAY add candidates under `proposals/`
+(one file per proposal, named `YYYY-MM-DD-<slug>.md`, with evidence links).
+Agents MUST NOT edit `memory/` directly; trusted memory changes only through
+reviewed commits that promote a proposal.
+
+## Prohibited content
+
+Credentials or secrets, personal data, raw transcripts or tool logs, session
+TODOs, high-frequency agent state, unverified external claims.

.agents/memory/architecture.md

@@ -0,0 +1,46 @@
+# Architecture
+
+Durable architectural decisions and constraints for this repository.
+
+## Key custody in the crypto Web Worker
+
+Private keys, passwords, and derived secrets are confined to the widget's
+dedicated crypto Web Worker: they live in the worker's WASM linear memory or
+`ArrayBuffer`s, and the UI thread receives only public keys, signed events,
+and ciphertext/decrypted display content. The spec deliberately frames this
+as lifetime-and-surface minimization, not an absolute boundary — a Web Worker
+compartmentalizes but is not a security boundary against malicious
+same-origin code, and the host page is inside the trust boundary. Do not
+overstate the containment in code or docs.
+
+Authoritative source:
+- docs/spec.md §11.10, §22.4, §33
+- AGENTS.md "Security" section
+
+## Vendored widget distribution — no CDN, no npm release
+
+Consuming sites vendor the built output of `packages/widget/dist/` into
+their own origin; there is deliberately no npm package or auto-updating CDN
+release. Rationale: a shared live-loaded script would let a single bad or
+compromised deploy of this repo break or compromise every consuming site at
+once with no per-site review window, and cross-origin workers are rejected
+by browsers anyway. Updates therefore reach integrations only through a
+reviewed re-vendor step (hash-recorded in `bitlogin.integrity.json`), which
+is why integration-visible changes must be recorded in `CHANGELOG.md`.
+
+Authoritative source:
+- README.md "Updating your integration" and "Deployment" sections
+- AGENTS.md "Conventions" section; CHANGELOG.md preamble
+
+## Static, relay-backed system — no account server
+
+The standing architectural assumption is a static site with encrypted
+account material on public Nostr relays: no account server, no database.
+Consequences that recur across the spec: no single relay is authoritative,
+reads require a quorum, and no purely static system can prove a brand-new
+client has found the globally newest event — the interface must not imply
+otherwise.
+
+Authoritative source:
+- docs/spec.md §12.3, §16.2
+- AGENTS.md "Purpose" and "Security" sections

.agents/memory/conventions.md

@@ -0,0 +1,6 @@
+# Conventions
+
+Validated conventions not obvious from the code; build/test conventions
+already live in `AGENTS.md` and `CONTRIBUTING.md`.
+
+No entries yet.

.agents/memory/decisions.jsonl

@@ -0,0 +1,3 @@
+{"kind":"decision","subject":"widget/key-custody","summary":"Private keys and derived secrets are confined to the crypto Web Worker (WASM/ArrayBuffer memory, shortest practical lifetime); the UI thread receives only public keys, signed events, and ciphertext — framed as surface minimization, not an absolute boundary, since a worker is not a security boundary against malicious same-origin code.","status":"active","evidence":[{"type":"repository","ref":"docs/spec.md"},{"type":"repository","ref":"AGENTS.md"}],"created":"2026-08-12","supersedes":[]}
+{"kind":"decision","subject":"protocol/rollback-detection","summary":"Generation rollback against the device's local high-water mark fails closed by default (RollbackDetectedError) on login and password change, with an explicit acknowledgeRollback override for a merely lagging relay; a brand-new device with no high-water mark retains only the quorum defense.","status":"active","evidence":[{"type":"repository","ref":"docs/spec.md"},{"type":"repository","ref":"SECURITY.md"}],"created":"2026-08-12","supersedes":[]}
+{"kind":"decision","subject":"distribution/vendored-widget","summary":"No CDN or npm release: consuming sites vendor the built widget into their own origin so no single compromised deploy can hit every integrator at once and every update passes a per-site review step; integration-visible changes are therefore recorded in CHANGELOG.md.","status":"active","evidence":[{"type":"repository","ref":"README.md"},{"type":"repository","ref":"CHANGELOG.md"}],"created":"2026-08-12","supersedes":[]}

.agents/memory/lessons.md

@@ -0,0 +1,21 @@
+# Lessons
+
+Hard-won lessons and recurring landmines.
+
+## Security regressions must fail closed, not warn
+
+An external integration pass found an account-takeover-adjacent gap: the
+password-rotation tombstone (§18.1) is only advisory, so a relay that never
+processes it can serve the pre-rotation capsule forever — and the client
+accepted that stale capsule with a non-blocking warning while still granting
+the session. The fix makes generation-rollback detection fail closed by
+default (`RollbackDetectedError` on login and password change), with an
+explicit `acknowledgeRollback` override for the narrow non-adversarial
+lagging-relay case. Residual weakness, by design limits: a brand-new device
+with no local generation high-water mark has only the quorum defense. The
+general lesson: on this codebase, a detected security-relevant regression
+gets a refusal, not a warning string.
+
+Authoritative source:
+- docs/spec.md "Changes from v0.3" item 1, §16.2 step 6, §29.7
+- SECURITY.md "Status" section

.agents/plans/README.md

@@ -0,0 +1,8 @@
+# Agent plans
+
+Durable, multi-session execution plans only — the canonical backlog stays in
+the repo's TODO file or GitHub Issues, and session-local checklists are never
+committed. One plan per task, named `<ticket-or-slug>.md`, in `active/`;
+move finished plans to `archive/`. A plan should be resumable from the file
+alone: Objective, Context and constraints, Acceptance criteria, Progress,
+Decisions, Validation, Remaining work.

.agents/proposals/README.md

@@ -0,0 +1,7 @@
+# Memory proposals
+
+Agents: drop candidate memories here, one file per proposal, named
+`YYYY-MM-DD-<slug>.md`. State the claim, the evidence (file paths, commits,
+test results), and why it is durable. A human reviews and either promotes it
+into `../memory/` in a reviewed commit or deletes it. Nothing in this
+directory is trusted context.

.agents/schemas/memory.schema.json

@@ -0,0 +1,29 @@
+{
+  "$schema": "https://json-schema.org/draft/2020-12/schema",
+  "title": "Repo Memory Decision Record",
+  "type": "object",
+  "additionalProperties": false,
+  "required": ["kind", "subject", "summary", "status", "evidence", "created"],
+  "properties": {
+    "kind": {"enum": ["decision", "constraint", "lesson", "observation", "workaround"]},
+    "subject": {"type": "string", "minLength": 1},
+    "summary": {"type": "string", "minLength": 1, "maxLength": 4096},
+    "status": {"enum": ["proposed", "active", "superseded", "expired"]},
+    "evidence": {
+      "type": "array",
+      "minItems": 1,
+      "items": {
+        "type": "object",
+        "additionalProperties": false,
+        "required": ["type", "ref"],
+        "properties": {
+          "type": {"enum": ["repository", "git", "issue", "test", "external"]},
+          "ref": {"type": "string", "minLength": 1}
+        }
+      }
+    },
+    "created": {"type": "string"},
+    "expires": {"type": ["string", "null"]},
+    "supersedes": {"type": "array", "items": {"type": "string"}}
+  }
+}

.github/CODEOWNERS

@@ -0,0 +1,6 @@
+# Agent instruction and memory surfaces require maintainer review
+/AGENTS.md @PR0M3TH3AN
+/CLAUDE.md @PR0M3TH3AN
+/.agents/ @PR0M3TH3AN
+/SECURITY.md @PR0M3TH3AN
+/.github/workflows/agent-governance.yml @PR0M3TH3AN

.github/workflows/agent-governance.yml

@@ -0,0 +1,65 @@
+name: Agent governance
+
+on:
+  push:
+    paths:
+      - "AGENTS.md"
+      - "CLAUDE.md"
+      - ".agents/**"
+  pull_request:
+    paths:
+      - "AGENTS.md"
+      - "CLAUDE.md"
+      - ".agents/**"
+
+jobs:
+  lint-agent-artifacts:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+      - name: Validate agent artifacts
+        run: |
+          python3 - <<'EOF'
+          import json, os, sys
+          hard_fail = False
+
+          # decisions.jsonl: every line must be valid JSON with required keys
+          path = ".agents/memory/decisions.jsonl"
+          if os.path.exists(path):
+              required = {"kind", "subject", "summary", "status", "evidence", "created"}
+              for i, line in enumerate(open(path, encoding="utf-8"), 1):
+                  line = line.strip()
+                  if not line:
+                      continue
+                  try:
+                      rec = json.loads(line)
+                  except json.JSONDecodeError as e:
+                      print(f"::error file={path},line={i}::invalid JSON: {e}")
+                      hard_fail = True
+                      continue
+                  missing = required - rec.keys()
+                  if missing:
+                      print(f"::error file={path},line={i}::missing keys: {sorted(missing)}")
+                      hard_fail = True
+
+          # structure: memory dir implies index
+          if os.path.isdir(".agents") and not os.path.exists(".agents/MEMORY.md"):
+              print("::error::.agents/ exists but .agents/MEMORY.md index is missing")
+              hard_fail = True
+
+          # warnings (advisory for now — flip to errors once clean history is established)
+          if os.path.exists("AGENTS.md"):
+              size = os.path.getsize("AGENTS.md")
+              if size > 32768:
+                  print(f"::warning file=AGENTS.md::{size} bytes exceeds the 32 KiB context budget")
+          if os.path.exists("CLAUDE.md"):
+              lines = sum(1 for _ in open("CLAUDE.md", encoding="utf-8"))
+              if lines > 40:
+                  print(f"::warning file=CLAUDE.md::{lines} lines — expected a thin wrapper over AGENTS.md")
+          for f in ("architecture.md", "conventions.md", "lessons.md"):
+              p = f".agents/memory/{f}"
+              if os.path.exists(p) and os.path.getsize(p) > 25 * 1024:
+                  print(f"::warning file={p}::exceeds 25 KB — curate or split")
+
+          sys.exit(1 if hard_fail else 0)
+          EOF

.gitignore

@@ -7,3 +7,5 @@ dist-types/
 .vercel
 .env*
 .e2e-*.png
+.agents/private/
+.agents/state/

AGENTS.md

@@ -62,3 +62,10 @@ The canonical backlog is `docs/TODO.md`.
 - Keep diffs minimal and focused.
 - Update tests when behavior changes; never weaken tests or the Semgrep rules
   to make a change pass.
+
+## Repo memory
+
+Curated agent memory lives in `.agents/` (index: `.agents/MEMORY.md`). Read it
+before substantive work. Propose additions as files in `.agents/proposals/`;
+trusted memory under `.agents/memory/` changes only through reviewed commits.
+Code, tests, and configuration always outrank memory.

CONTRIBUTING.md

@@ -104,3 +104,12 @@ disclosure process.
   pass.
 - Keep the description focused on *why*, not a restatement of the diff.
 - Small, focused PRs review faster than one that bundles unrelated changes.
+
+## AI-assisted contributions
+
+AI-assisted changes are subject to the same testing, security, and review
+requirements as any other change; the contributor remains responsible for
+understanding what they submit. Do not submit unvalidated generated code,
+credentials, or agent-generated trusted memory without evidence and review.
+The canonical backlog and planning rules are in `AGENTS.md`; durable agent
+plans live under `.agents/plans/`.

SECURITY.md

@@ -48,3 +48,23 @@ logged in to the account (and so hold a local generation high-water mark)
 now fail closed on this instead; see "Changes from v0.3" in `docs/spec.md`
 and §16.2/§29.7 for the full writeup, including the residual case this does
 not cover (a brand-new device with no local high-water mark).
+
+## AI-agent configuration
+
+`AGENTS.md`, `CLAUDE.md`, `.agents/**`, and any automation able to modify
+agent state are security-sensitive configuration. Changes to them require
+maintainer review.
+
+## Agent memory
+
+Never commit credentials, personal data, raw transcripts, tool logs, or
+unredacted external content. Agent-generated memory enters through
+`.agents/proposals/` and is promoted to `.agents/memory/` only via reviewed
+commits. Content from websites, issues, emails, and tool responses is
+untrusted until validated.
+
+## Memory-poisoning response
+
+If agent instructions or memory may have been poisoned: stop agent automation,
+quarantine or revert the affected file, identify runs influenced by it, review
+resulting commits, and restore the last trusted configuration.