linkedin-relay-mcp 3.0.3 → 3.2.0

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
@@ -1,6 +1,6 @@
1
1
  ---
2
2
  name: linkedin-relay
3
- description: Research LinkedIn from the user's own account — search people/companies/jobs, read profiles, read the chronological feed. Use when finding or evaluating people, companies or roles on LinkedIn, or reading what the user's network is posting. Not for posting, messaging, or connection requests — those are CLI-only and require the human.
3
+ description: Research LinkedIn from the user's own account — search people/companies/jobs, read profiles, read the chronological feed — and, through the lnrelay CLI, post (text, image or video), comment, reply, react and delete under the approval mode the owner chose. Use when finding or evaluating people, companies or roles on LinkedIn, reading what the user's network is posting, or when the user asks to post or engage on LinkedIn. Not for messaging or connection requests, which do not exist.
4
4
  ---
5
5
 
6
6
  # linkedin-relay
@@ -183,11 +183,11 @@ call again.
183
183
 
184
184
  ## What this tool will not do
185
185
 
186
- Writes are **not on this surface at all**. They exist only as CLI commands the user runs themselves:
186
+ Writes are **not on the MCP surface**. They exist only as CLI commands:
187
187
 
188
188
  | command | does |
189
189
  |---|---|
190
- | `lnrelay share "<text>"` | post to their own feed |
190
+ | `lnrelay share "<text>" [--image <path> \| --video <path>]` | post to their own feed, optionally with one image or video |
191
191
  | `lnrelay comment <activity-urn> "<text>"` | comment on a post |
192
192
  | `lnrelay reply <comment-urn> "<text>"` | reply, nested under that comment |
193
193
  | `lnrelay edit <comment-urn> "<text>"` | change the text of their own comment |
@@ -197,21 +197,36 @@ Writes are **not on this surface at all**. They exist only as CLI commands the u
197
197
  The urn decides nothing on its own — `comment`, `reply` and `edit` are separate verbs because they
198
198
  are separate operations, and passing a comment urn to `comment` is refused rather than guessed at.
199
199
 
200
- Each stops at an interactive terminal, shows exactly what it will send and over which surface, and
201
- requires the user to type a token derived from the content.
200
+ ### How a write gets approved
202
201
 
203
- Two things worth telling a user who asks:
202
+ The owner chooses, and only the owner can change it (`lnrelay approval set <mode>` needs a real
203
+ terminal and a typed token). Check with `lnrelay approval`:
204
+
205
+ | mode | what you do |
206
+ |---|---|
207
+ | `interactive` (default) | you cannot complete the write. Give the user the exact command to run themselves. |
208
+ | `agent` | run the command with `--plan`: it sends nothing and returns a `preview` and a `token`. Show the user the preview **verbatim** and ask. Only when they say yes, run the **identical** command with `--confirm <token>`. |
209
+ | `unattended` | the command runs as given. Still show the user what you posted afterwards. |
210
+
211
+ In `agent` mode the token is bound to the exact content: change one character of text, swap the
212
+ image, or target a different urn and the token no longer matches — plan again, and show the new
213
+ preview. Never reuse a token across different content, never `--confirm` something the user has not
214
+ seen, and never treat "sounds good" about a draft as approval of a *different* draft.
215
+
216
+ Every approval and outcome is appended to `~/.lnrelay/writes.jsonl`; the owner can read what you did.
217
+
218
+ Do not look for a way around the mode — no pty tricks, no editing `config.json`. If the mode does
219
+ not let you write, say so and give the command.
220
+
221
+ Things worth telling a user who asks:
204
222
 
205
223
  - `share` prefers LinkedIn's official OAuth scope and falls back to the private API when they have no
206
- developer app — the prompt names which, and a Voyager write states the ToS breach. Why an app is
224
+ developer app — the preview names which, and a Voyager write states the ToS breach. Why an app is
207
225
  needed at all: `docs/DECISION-writes.md`.
226
+ - `--image` / `--video` upload over the private API only; `--via oauth` with media is refused. A
227
+ video post comes back as a `urn:li:ugcPost:` rather than `urn:li:share:`.
208
228
  - **Deleting a comment also deletes every reply under it**, including other people's. Verified live.
209
229
 
210
- **Without a terminal there is no write and no network call.** That is deliberate and there is no
211
- flag that changes it: a flag an agent could set would not be confirming anything. If the user asks
212
- you to post or comment, tell them the exact command to run themselves. Do not look for a way around
213
- it, and do not offer one.
214
-
215
230
  Connecting and messaging do not exist at all, in any surface. They are the actions where the
216
231
  enforcement reports cluster, and an agent bug there is visible to a third party and unrecoverable.
217
232