docs: add 2026-06-01 AgentKeys progress report

CHANGELOG.md

@@ -0,0 +1,19 @@
+# Changelog
+
+All notable changes to AgentKeys are tracked here.
+
+## 2026-06-01
+
+### Added
+
+- Full agent bootstrap ceremony for HDKD-derived agents, including broker link-code endpoints, daemon-side redeem, pending bindings, and the bind/grant split. See PR #149.
+- Hooks-first Hermes wiring for AgentKeys, including `agentkeys wire`, `agentkeys hook check`, `agentkeys hook audit`, `agentkeys hook memory-inject`, and the operator runbook. See PR #141.
+
+### Changed
+
+- Locked the memory architecture direction around the gated-backend model: AgentKeys owns the encrypted, per-actor store and deterministic gate, while ranking and extraction engines stay pluggable. See PR #146.
+- Cleaned and indexed the architecture docs around memory, universal gate behavior, and design-record links. See PR #146.
+
+### Notes
+
+- The next active slices are namespace-bound memory caps, device lifecycle handling, hosted-LLM MCP deployment, and the brand asset PR.

docs/reports/2026-06-01/article.md

@@ -0,0 +1,27 @@
+# AgentKeys Daily Progress, 2026-06-01
+
+Status: draft for review
+Image: `assets/article-illustration.png`
+
+Three quiet pieces moved AgentKeys forward this weekend.
+
+The first is safer agent pairing. The agent now gets its own device key inside its own environment. The master approves it. That sounds small, but it is the product line in one sentence: give the agent a scoped identity without handing it the master key.
+
+The second is memory. The decision is now sharper: AgentKeys should own the encrypted memory store and the gate around it. The ranking engine can stay pluggable. The durable part is the user's memory and access policy, and that belongs under the authority layer.
+
+The third is Hermes. The hook path turns AgentKeys from a tool an agent may call into a gate the host runs. Permission checks, audit writes, and memory injection move into the lifecycle of the agent runtime. That is the difference between asking an agent to be careful and making the boundary part of the system.
+
+Plainly: this is still build-in-public work. AgentKeys is not being described as finished infrastructure. The important thing is that the shape is getting narrower and more useful. Broker credentials, do not proxy every action. Gate memory, do not dump the whole store into context. Bind agents as first-class devices, do not treat them as loose scripts with `.env` files.
+
+What comes next is already visible. Namespace-bound memory caps are in review. Device lifecycle work has been split into local unbind and on-chain self-revocation. The hosted-LLM path is now parked in its own issue so the team can stay focused on the local Task-Host route first.
+
+Same thesis, tighter surface. AgentKeys is becoming the authority layer between AI agents and the things they are allowed to know, access, and do.
+
+## Internal Source Notes
+
+- PR #149 merged the full agent bootstrap ceremony.
+- PR #146 merged the gated memory decision and universal gate pattern.
+- PR #141 merged the Hermes hooks-first flow.
+- PR #150 is open for namespace-bound memory caps.
+- Issues #155 and #156 split device lifecycle into on-chain self-revocation and local unbind/re-pair.
+- Issue #152 parks the hosted-LLM MCP endpoint path.

docs/reports/2026-06-01/run-summary.md

@@ -0,0 +1,54 @@
+# AgentKeys Daily Automation Run, 2026-06-01
+
+Status: generated for review
+
+## GitHub Triage
+
+Updated issue labels, milestones, and AgentKeys Features project fields:
+
+| Issue | Labels | Milestone | Project fields |
+|---|---|---|---|
+| #147 | enhancement, area/memory, area/mcp | M2 | Size L, Kind Feature, Priority High, Phase v1, Risk Medium |
+| #152 | enhancement, area/mcp, area/infra | M3 | Size L, Kind Feature, Priority Low, Phase v2, Risk Medium |
+| #153 | enhancement, area/audit, area/ui | M2 | Size M, Kind Feature, Priority Medium, Phase v1, Risk Medium |
+| #155 | enhancement, area/identity, area/signer, area/daemon, area/cli | M4 | Size XL, Kind Security, Priority Medium, Phase v2, Risk High |
+| #156 | enhancement, area/identity, area/daemon, area/cli | M4 | Size M, Kind Feature, Priority High, Phase v1, Risk Medium |
+
+## Issue Backlinks
+
+Added merged-PR status comments to:
+
+- #103 for the merged Hermes hooks and HDKD bootstrap progress.
+- #133 for PR #141 and the remaining runtime-adapter work.
+- #144 for PR #149 closing the full bootstrap ceremony.
+- #147 for PR #146 locking the memory store/gate direction and PR #150 opening the first implementation slice.
+
+## Merged PRs Used For Today
+
+- #149, merged 2026-05-31: full agent bootstrap ceremony.
+- #146, merged 2026-05-31: memory build-vs-gate decision and universal gate pattern.
+- #141, merged 2026-05-31: Hermes hooks-first wire flow.
+
+## Open PRs Noted
+
+- #150: namespace-bound memory caps.
+- #154: lowercase GHCR image tag fix for the MCP server publish job.
+- #157: AgentKeys logo assets. This is still draft/open, so today's images do not invent or duplicate a logo.
+
+## Drafts Created
+
+- X post: `docs/reports/2026-06-01/x-post.md`
+- Article: `docs/reports/2026-06-01/article.md`
+- X image: `docs/reports/2026-06-01/assets/x-post-image.png`
+- Article illustration: `docs/reports/2026-06-01/assets/article-illustration.png`
+
+## Voice Inputs
+
+The drafts follow the Heima/AgentKeys voice from:
+
+- `~/Projects/heima-market/marketing-os/knowledge/heima/messaging.md`
+- `~/Projects/heima-market/marketing-os/projects/heima/persona.md`
+- `~/Projects/heima-market/marketing-os/knowledge/agentKeys/messaging.md`
+- `~/.claude/skills/agentkeys-write/voice-samples/2026-04-24-introducing-agentkeys.md`
+
+Applied constraints: engineer-direct, mechanism-first, no hype, no public Litentry references in public copy, no token-selling language, no unsupported security claims, and no HN draft.

docs/reports/2026-06-01/x-post.md

@@ -0,0 +1,25 @@
+# AgentKeys X Post Draft
+
+Status: draft for review
+Date: 2026-06-01
+Image: `assets/x-post-image.png`
+
+## Post
+
+AgentKeys shipped 3 quiet pieces this weekend: safer agent pairing, a clearer memory gate, and the first Hermes hook path.
+
+Translation: agents get scoped access without holding the keys themselves.
+
+Broker, not proxy. Gate, not hope.
+
+## Image Direction
+
+Use `assets/x-post-image.png`.
+
+Style note: clean technical editorial illustration, light background, charcoal linework, soft cyan and restrained green accents. No text, no logo, no token imagery.
+
+## Internal Source Notes
+
+- PR #149 merged the HDKD agent bootstrap and broker link-code flow.
+- PR #146 locked the memory store/gate decision.
+- PR #141 merged the Hermes hooks-first wire flow.