## Intel — 2026-04-16 (ElizaOS)

### Data availability / confidence
- **Daily artifacts for 2026-04-16 not found** (`2026-04-16.md` missing). This intel is based on:
  - **Monthly rollup (2026-04-01 → 2026-05-01)** for `elizaos/eliza`
  - **Weekly summary (Apr 5–11, 2026)** and linked PR/issue content
- Confidence is **medium** for trend and prioritization, **low** for “what happened today” specifics.

---

## 1) Data Pattern Recognition (velocity, engagement, adoption, pain-point correlation)

### Development velocity (April-to-date rollup)
**Repo: `elizaos/eliza` (month interval)**
- PRs: **4 opened**, **3 merged**
- Issues: **7 opened**, **1 closed**
- Contributors: **14 active**
- Code churn: **+6,785 / -3,732** across **81 files**, **70 commits**
- Interpretation:
  - Throughput is decent, but **issue inflow > closure** suggests backlog pressure.
  - Churn is concentrated in **core runtime/message pipeline**, implying elevated regression risk.

### Where changes are clustering (hot zones)
- **Message pipeline / shouldRespond / group routing**
  - PRs touching decision logic and group behavior: **#6712**, **#6562**, **#6709**
- **DevEx harness & workspace/submodule workflow**
  - PR **#6702** merged: large structural changes (agent harness + plugin submodules)
- **Economic layer & trust**
  - Weekly emphasis: wallet plugin, cross-chain ops, cryptographic identity (“AgentID”), capability authorization discussion.
  - Issues show **strong community pull** for marketplace + safety + authorization layers.

### Community engagement patterns (what people are asking for)
Open “top issues” indicate demand clusters:
1. **Marketplace / economic interoperability**: MAXIA marketplace plugin proposal (**#6700**)
2. **Capability authorization & auditability**: SINT capability token enforcement (**#6707**)
3. **Safety tooling for on-chain action**: SafeAgent token safety checks (**#6706**)
4. **Bootstrapping reliability / install friction**: `elizaos create` failing on macOS due to bun dependency (**#6704**)
5. **Incentives / contribution economy narrative**: AIGEN Protocol (**#6708**)

**Signal:** Community is aligned with the project’s pivot to “trust-based agent economy,” but **onboarding and correctness** in core runtime remain the most immediate adoption blockers.

### Feature adoption signals (proxy-based)
No direct usage telemetry provided; infer adoption pressure from:
- Bugs found while testing real connectors (Discord/milady) → PR **#6709** exists specifically due to connector-driven failures.
- Dev harness merged (**#6702**) suggests internal team prioritizing “boot repo quickly,” usually driven by onboarding friction.

### Pain point correlation across channels
- **Tool execution reliability** (toon params, async action continuation loops) + **group-room routing** + **memory gating/logging** all converge on a single theme:
  - “Agents either act incorrectly (wrong/no params), speak when they shouldn’t (loops/group routing), or lose/skip memory unexpectedly.”
- Onboarding pain (`elizaos create` + bun postinstall) correlates with the repo investing in dev harness—both indicate **setup fragility** is limiting contributor growth.

---

## 2) User Experience Intelligence (impact/themes, usage vs intended, opportunities, sentiment)

### Feedback themes (categorized by impact)

**High impact (blocks adoption / causes visible failures)**
- **Project creation fails on macOS** due to bun packaging behavior (**Issue #6704**)
  - Outcome: user gets *no project directory* (CLI cleans up), which is a high-friction “dead end.”
- **Connector action execution failures (TOON encapsulation)** due to missing schema `params` (**PR #6709**)
  - Prior behavior: required params never supplied → actions silently fail or underperform.
- **Group-room misrouting / ping-pong loops risk** (PR **#6712** changes shouldRespond prompts + deterministic routing)
  - This directly impacts perceived intelligence and “social correctness” of agents.

**Medium impact (trust, debuggability, operator control)**
- **Memory creation gating + observability** (PR **#6562** merged)
  - Benefits: audit logs, provider timeouts, memory controls.
  - Risks flagged in review: specific logic bugs (IGNORE persistence with allowlist; zero-vector embedding fallback) could silently undermine trust in memory.

**Strategic UX (value expansion, but must not degrade safety)**
- Plugin proposals: marketplace, safety scoring, economic memory, capability tokens.
  - These are attractive, but increase blast radius unless paired with clear authorization + safe defaults.

### Usage patterns vs intended design (mismatches)
- **Non-streaming connectors** (Discord, milady) rely on TOON schema completeness. Missing schema fields led to “LLM wasn’t asked, so it didn’t output it.”
  - Implementation lesson: prompt/docs ≠ runtime schema; UX depends on schema parity.
- **Async task actions** (PTY-backed coding/orchestration) were treated as non-terminal, causing repeated continuation responses (“filler spam”).
  - Intended: single clean result; observed: loop noise → user distrust.

### Implementation opportunities (concrete)
1. **Fail-forward onboarding**
   - CLI should not delete the project directory on build failure; preserve artifacts + print exact remediation.
2. **Schema/prompt contract tests**
   - Add automated tests ensuring prompt templates’ documented fields are present in runtime schemas (prevents repeats of TOON params gap).
3. **Safety-by-default action gating**
   - Given incoming marketplace/swap tooling, first-class “capability-based authorization” should be surfaced as a core pattern (even if implemented via plugin).

### Community sentiment (inferred)
- **Positive pull**: many plugin proposals, incentives, security/safety tooling—community wants to build.
- **Fragility concern**: multiple reviews flag correctness risks in core. If regressions land, sentiment can shift quickly from “excited builders” to “unstable framework.”

---

## 3) Strategic Prioritization (impact vs risk, dependencies, resource allocation)

### Priority matrix (next 1–2 weeks)

#### P0 — Must fix (user impact high, risk high if left unresolved)
1. **`elizaos create` macOS failure (bun dependency/postinstall)** — **Issue #6704**
   - Impact: hard blocks new users on macOS (large share of devs).
   - Recommendation: remove `bun` runtime dependency; use `@types/bun` only; add install-time check and actionable error message.

2. **Core correctness regressions in merged runtime hardening PR (#6562)**
   - Two high-risk review findings to confirm/patch quickly:
     - **IGNORE memory persistence allowlist bug** (allowedSources includes `"agent_response"` hardcoded)
     - **Zero-vector embedding fallback** corrupting memory retrieval
   - Impact: silent correctness failures → hardest to diagnose → damages trust.

#### P1 — Should merge/finalize with targeted fixes (high impact, manageable risk)
3. **TOON params + async continuation loop fix** — **PR #6709**
   - If not merged yet, prioritize merge after quick regression pass; it directly fixes connector UX and reduces spam.
   - Add a regression test: “required params action via TOON connector path.”

4. **Group addressee routing + anti-loop guidance** — **PR #6712**
   - Blocker: **aliasEntity ambiguity bug** (breaks name resolution when `entityId ≠ agentId`).
   - Action: fix bug + add test covering aliasEntity behavior; then merge.
   - Outcome: reduces multi-agent room chaos; improves “social UX.”

#### P2 — Gate behind safety/architecture (strategic, but avoid premature integration)
5. **plugin-mnemopay (economic memory)** — **PR #6701**
   - Current review flags: no persistence, NaN config corruption, null deref risks, unbounded growth, action validation conflicts, no tests.
   - Recommendation: do **not** merge until:
     - persistence strategy exists (SQL adapter or file-backed store),
     - strict config validation,
     - bounded memory + eviction,
     - action triggers disambiguated,
     - minimal test suite added.

6. **Marketplace / safety / capability-token proposals** — Issues **#6700, #6706, #6707**
   - Dependency: clarify/ship **AgentID + capability authorization** direction (even as “reference plugin + spec”).
   - Recommendation: create an “Approved plugin capability levels” guideline:
     - read-only (T0), act (T2), commit (T3), with examples and default prompts.

---

## Quantitative “watch list” (leading indicators)
- Backlog pressure: **7 new issues vs 1 closed** (monthly rollup)
- Core churn risk: **81 files changed**, with multiple PRs touching `message.ts` / `runtime.ts`
- Contributor concentration: top contributor “odilitime” dominates score → **key-person risk**
- Plugin inflow: at least **3–5 substantive plugin proposals** tied to financial actions → safety expectations rising.

---

## Actionable Recommendations (optimize dev efficiency + UX)

### A) Stabilize onboarding and reduce first-run failures (P0)
- Fix #6704 (bun dependency) and add:
  - CLI preflight checks: bun version, postinstall state, and clear remediation.
  - Change behavior: **do not delete** scaffolded project on build failure; leave directory + `elizaos doctor` output.

### B) Add “contract tests” for message formats and connector paths (P1)
- Introduce tests that validate:
  - TOON schema contains fields documented in prompts (`params`, attachments, etc.).
  - Non-streaming connector path successfully executes an action with required params.
- Outcome: reduces regressions that only appear in Discord/Telegram adapters.

### C) Close correctness gaps in memory + embeddings (P0)
- Patch and backfill:
  - If embedding fails: do **not** persist a zero vector; either retry, mark as “embedding_pending,” or store without vector and exclude from semantic search gracefully.
  - Fix IGNORE persistence allowlist logic to match intended design (agent responses shouldn’t require a source allowlist token users will never add).
- Add a small migration/cleanup script to identify zero-vector memories if already stored.

### D) Safety-by-default for economic actions (P1/P2)
- Before endorsing marketplace/swap plugins:
  - publish a minimal “capability approval flow” reference (core doc + example plugin),
  - define a standard event/audit format (action, params hash, approvals, chain/network, txid).
- This aligns with AgentID direction and prevents ecosystem fragmentation.

### E) Reduce coordination risk (team efficiency)
- Given contributor concentration, define explicit ownership:
  - Runtime/message pipeline owner
  - Dev harness/workspaces owner
  - Trust/economic layer owner (AgentID + wallet + authorization)
- Ensure each major PR has a “rollout checklist” (latency, connector tests, memory integrity checks).

---

## Critical Path Dependencies (next)
1. **Onboarding reliability (CLI + bun)** → directly impacts contributor growth and plugin ecosystem throughput.
2. **Core runtime correctness (memory, embeddings, shouldRespond/group routing)** → foundational trust for economic layer.
3. **Capability authorization standard (spec + reference implementation)** → prerequisite for safely scaling financial/marketplace plugins.

---

## Immediate next-step checklist (24–72 hours)
- [ ] Triage and patch **#6704** (bun runtime dependency) with a release note and CLI behavior change (preserve directory).
- [ ] Audit merged **#6562** for the two flagged correctness issues; ship hotfix PR(s).
- [ ] If not already: merge **#6709** after confirming no regressions in XML path and streaming connectors.
- [ ] Fix **#6712** aliasEntity ambiguity + add missing test; then proceed with merge to reduce group-room UX failures.
- [ ] Put **#6701** (mnemopay) behind a “needs persistence + tests” gate; offer a concrete persistence pattern (SQL adapter integration) to accelerate contributor success without merging fragile code.