bereach-openclaw 0.2.16 → 0.2.17

package/openclaw.plugin.json

@@ -1,7 +1,7 @@
 {
   "id": "bereach",
   "name": "BeReach",
-  "version": "0.2.16",
+  "version": "0.2.17",
   "description": "LinkedIn outreach automation — 33 tools, auto-reply commands",
   "configSchema": {
     "type": "object",

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "bereach-openclaw",
-  "version": "0.2.16",
+  "version": "0.2.17",
   "description": "BeReach LinkedIn automation plugin for OpenClaw",
   "license": "AGPL-3.0",
   "scripts": {

package/skills/bereach/SKILL.md

@@ -1,7 +1,7 @@
 ---
 name: bereach
 description: "Automate LinkedIn outreach via BeReach (berea.ch). Use when: prospecting, engaging posts, scraping engagement, searching LinkedIn, managing inbox, running campaigns, managing invitations. Requires BEREACH_API_KEY."
-lastUpdatedAt: 1772673508
+lastUpdatedAt: 1772678140
 metadata: { "openclaw": { "requires": { "env": ["BEREACH_API_KEY"] }, "primaryEnv": "BEREACH_API_KEY" } }
 ---
 
@@ -17,7 +17,7 @@ Automate LinkedIn prospection and engagement via BeReach.
 
 | Sub-skill             | Keywords                                                                                                                  | URL                                                     | lastUpdatedAt |
 | --------------------- | ------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------- | ------------- |
-| Lead Magnet           | lead magnet, comment to DM, resource delivery, post giveaway, auto-accept invitations, scheduled run, recap, campaign stats, pause | sub/lead-magnet.md       | 1772673508    |
+| Lead Magnet           | lead magnet, comment to DM, resource delivery, post giveaway, auto-accept invitations, scheduled run, recap, campaign stats, pause | sub/lead-magnet.md       | 1772678140    |
 | OpenClaw Optimization | openclaw, model, opus, sonnet, haiku, config, SOUL.md, heartbeat, prompt caching, AI cost reduction, /model               | openclaw-optimization.md | 1772619338    |
 | SDK Reference         | sdk, method, parameter, signature, reference, api, script                                                                 | sdk-reference.md         | 1772672714             |
 
@@ -81,9 +81,13 @@ The SDK auto-reads `BEREACH_API_KEY` from the environment. NEVER hardcode tokens
 
 - Lead magnet scripts use the SDK. Tools are for interactive agent use only.
 - Import from the `bereach` SDK exclusively. If a method doesn't exist, the operation doesn't exist.
-- Scripts MUST be TypeScript (`.ts`). Run with `npx tsx script.ts`.
-- Run `npx tsc --noEmit` before executing to catch type errors.
-- Load the [SDK Reference](sdk-reference.md) sub-skill for method signatures.
+- Scripts MUST be TypeScript (`.ts`). Load the [SDK Reference](sdk-reference.md) sub-skill for method signatures.
+
+**Mandatory validation loop** — after generating or editing any script:
+
+1. Run `npx tsc --noEmit` in the script directory. If type errors: fix them using the SDK Reference, then repeat.
+2. Run `npx tsx <script-name>.ts` — fix runtime errors (e.g. "linkedinInbox is not defined", "visitProfile is not a function"). If config missing, run with minimal env to reach first SDK call.
+3. Only after both pass: consider the script ready.
 
 ### Always
 
@@ -139,7 +143,7 @@ More workflows coming soon. You can build your own using the SDK methods and too
 
 Crons are OpenClaw scheduled tasks. Create entries in `~/.openclaw/cron/jobs.json` (via `openclaw cron add`).
 
-**Lead Magnet** — 3 global crons. Config at `~/.openclaw/lead-magnet.json`. Each cron runs one script that loops over all enabled campaigns:
+**Lead Magnet** — 3 global crons. Campaign config at `~/.bereach/lead-magnet.json`. Each cron runs one script that loops over all enabled campaigns:
 
 ```json
 { "id": "lm-comments", "every": "1h", "skill": "bereach", "sessionTarget": "spawn", "prompt": "Run lm-comments.ts and report recap" }

package/skills/bereach/sub/lead-magnet.md

@@ -1,7 +1,7 @@
 ---
 name: bereach-lead-magnet
 description: "Lead magnet workflow — deliver a resource to everyone who engages with a LinkedIn post."
-lastUpdatedAt: 1772673508
+lastUpdatedAt: 1772678140
 ---
 
 # BeReach Lead Magnet Skill
@@ -16,6 +16,14 @@ Requires [BeReach SKILL.md](SKILL.md) (constraints, SDK setup, anti-hallucinatio
 
 All lead magnet scripts MUST be TypeScript using the `bereach` SDK (see main skill for SDK setup and rules).
 
+**Always use full SDK paths** — e.g. `client.linkedinChat.findConversation`, `client.linkedinScrapers.visitProfile`. Never infer the resource from method name alone (e.g. "inbox" ≠ linkedinInbox; "visit" ≠ linkedinActions.visitProfile).
+
+**Validation loop (mandatory)** — after writing each script:
+
+1. `npx tsc --noEmit` — fix any type errors using SDK Reference.
+2. `npx tsx <script-name>.ts` — fix any runtime errors (e.g. undefined method, wrong resource). If config is missing, run with minimal env to at least reach the first SDK call.
+3. Repeat until both pass. Only then save or paste the script.
+
 ## Tone
 
 All script output and chatbot messages are **for the end user**. Write like you're talking to a non-technical person. Never mention internal concepts like "Track A", "Layer 2", "DM guard", "server-side dedup", "retryAfter", "URN", "O(1) lookup", etc. The campaign slug is fine to show — it helps the user know which campaign is running. Just say what happened in plain language: "Sent the resource to 12 people", "Skipped 3 (already received it)", "Outside active hours, will resume at 7:00".
@@ -129,7 +137,7 @@ For each campaign (from config): scrape commenters and engage them.
 2. **Fetch**: `collectComments` with `campaignSlug` → response: `profiles[]` (NOT "comments"). Paginate: while `hasMore`, increment `start` by `count`.
 3. Each profile has: `profileUrl`, `commentUrn`, `name`, `profileUrn`, `actionsCompleted: { message, reply, like, visit, connect }`, `knownDistance`, `hasReplyFromPostAuthor`
 4. For each profile (skip own URN, skip `actionsCompleted.message === true`):
-   a. If `knownDistance` is set, use it (skip visit, saves 1 credit). Otherwise: `visitProfile` with `campaignSlug` → `memberDistance`, `pendingConnection`
+   a. If `knownDistance` is set, use it (skip visit, saves 1 credit). Otherwise: `client.linkedinScrapers.visitProfile` with `{ profile, campaignSlug }` → `memberDistance`, `pendingConnection`
    b. Distance 1:
    - **DM dedup** (see "DM dedup" section): check both layers. If either says "already sent", skip. Otherwise: `sendMessage` with `campaignSlug`.
    - Skip reply if `actionsCompleted.reply === true` or `hasReplyFromPostAuthor === true`. Otherwise: `replyToComment` with `commentUrn` and `campaignSlug`
@@ -147,7 +155,7 @@ Accept pending invitations and deliver the resource to campaign-related invitees
 - **Campaign attribution**: `listInvitations` returns ALL pending invitations. Cross-reference each inviter against campaign commenter lists to find which campaign (if any) they belong to. Accept all invitations, but only DM if campaign-matched.
 - For each invitation:
   1. `acceptInvitation` with `invitationId` and `sharedSecret`
-  2. If campaign matched: `visitProfile` with matched `campaignSlug`, then **DM dedup** → DM if not already sent.
+  2. If campaign matched: `client.linkedinScrapers.visitProfile` with `{ profile, campaignSlug }`, then **DM dedup** → DM if not already sent.
   3. If no campaign match (organic invitation): accept only, no DM.
 
 ### Script 3 — New connections
@@ -156,7 +164,7 @@ For each campaign (from config): check if distance 2+ commenters have now connec
 
 - Re-fetch commenters: `collectComments` with `campaignSlug` — paginate all
 - Filter: `actionsCompleted.message === false` and `knownDistance > 1` (skip profiles where `knownDistance` is null — they haven't been visited yet and belong to Script 1)
-- For each: `visitProfile` with `campaignSlug` → check `memberDistance` and `pendingConnection`
+- For each: `client.linkedinScrapers.visitProfile` with `{ profile, campaignSlug }` → check `memberDistance` and `pendingConnection`
   - `memberDistance === 1`: **DM dedup** (see "DM dedup" section) → DM the resource if not already sent.
   - `pendingConnection === "pending"`: not connected yet, skip
 
@@ -172,7 +180,7 @@ Check in order. If either says "already sent", skip the DM.
 
 **Layer 1 — `actionsCompleted.message`** (free, within campaign). Already checked in the script loop — `true` → skip.
 
-**Layer 2 — `/find` inbox lookup** (0 credits, cross-campaign). Only when `resourceLink` is set. Call `findConversation` with `includeMessages: true` → `{ found, messages }`. `messages` is at the **top level**, not nested. If any message text contains `resourceLink`: skip DM, call `syncActions` to mark done. If found but no link in messages: pass `messages` as context for DM tone adaptation. The function must return both skip decision and messages.
+**Layer 2 — DM guard** (0 credits, cross-campaign). Only when `resourceLink` is set. Call `client.linkedinChat.findConversation` with `{ profile, includeMessages: true }` → `{ found, messages }`. `messages` is at the **top level**, not nested. If any message text contains `resourceLink`: skip DM, call `client.campaigns.syncActions` to mark done. If found but no link in messages: pass `messages` as context for DM tone adaptation. The function must return both skip decision and messages.
 
 **Fail-safe**: if `findConversation` errors or all retries fail, **skip this profile** — do NOT send the DM. A failed lookup treated as "not found" causes duplicates. The profile will be retried next run.
 
@@ -206,7 +214,7 @@ Follow pacing rules from the main skill (Constraints #2). No exceptions — slee
 
 ## Cron
 
-Three global crons. Config at `~/.openclaw/lead-magnet.json`. Each cron runs one script that loops over all enabled campaigns.
+Three global crons. Campaign config at `~/.bereach/lead-magnet.json`. Each cron runs one script that loops over all enabled campaigns.
 
 All crons MUST use `"sessionTarget": "spawn"` to avoid blocking the main session. Each spawned sub-agent runs the script independently and pushes the recap back to the user when done.