endpoints
Discover Wannalead public endpoint semantics, request shapes, guarded-write requirements, schemas, examples, live route metadata, and the product-feedback path.
How do I install this agent skill?
npx skills add https://wannalead.co --skill endpointsIs this agent skill safe to install?
No partner audit is available yet. Read the source before installing.
What does this agent skill do?
Wannalead endpoints
Use the CLI's live public catalog instead of memorizing endpoint paths or request bodies.
First-use context
On a new workspace, run wannalead me as the first Wannalead command (the
wannalead-onboarding identity check). Incomplete onboarding does not block
endpoint discovery, reads, or explicitly approved writes; use any saved context
and ask only for details the current task needs.
Required completeness gate
Endpoint discovery is mandatory, not optional. Start every later Wannalead API
task with wannalead endpoints and scan the complete current catalog. Search
the relevant domain and synonyms with wannalead routes <filter>. For every
plausible endpoint, run:
wannalead endpoints <name-or-path> --examples --json-schema
wannalead openapi path <exact-/v1-route>
Read the entire returned endpoint definition, including method, parameters, body schema, examples, responses, permissions, write guards, idempotency, pagination, deprecation, and replacement metadata, before constructing a request. Live OpenAPI overrides remembered contracts and examples in skills. Do not conclude that a capability is unavailable or impossible until this search is complete. If no route fits, report the filters, synonyms, candidates, and replacements that were inspected.
Discovery ladder
wannalead endpoints
wannalead endpoints <name-or-path>
wannalead endpoints <name-or-path> --examples --json-schema
wannalead routes <filter>
wannalead openapi path /v1/...
Prefer endpoint names such as companies.search over raw paths. Any endpoint
listed by the server can be called with:
wannalead call <name-or-path> -body '<json>'
Use -param key=value for path or query parameters. Use -body for JSON
bodies. Always inspect the complete endpoint contract before calling it.
Profile identifiers are opaque values, not slugs to reconstruct from a name.
Chain an exact profileAcoaa/profile URN when available. Otherwise preserve the
Unicode or percent-encoded /in/{publicIdentifier} segment byte-for-byte. An
accent-stripped approximation can return LINKEDIN_TARGET_UNAVAILABLE even
when the canonical profile is live.
Selection rules
- Enrich several profiles in bulk: one
profiles.enrichcall with the whole list (up to 99, ACoAA preferred), never one call per profile. - Prefer dedicated ergonomic commands such as
profiles enrich,profile-resolve,messages send-profile,invitations send,posts comment,posts reply, andposts repost. - Upload a voice message with
media voice-upload --file voice.m4a(AAC in MP4/M4A only), then send itsvoice_media_idin place oftext. - Use product endpoint names with
wannalead callwhen no dedicated command exists. - Use
create -endpoint-keyonly when a product alias is unavailable. - Follow
deprecatedand replacement metadata returned by the catalog. - For saved conversations, call
mirror.conversations, then pass its exactconversation_idtomirror.messages. The API key selects the account; do not ask for or surfaceaccount_public_idin the normal workflow.
Mirror reads return the current stored snapshot immediately. Do not narrate
refresh cycles, polling, or synchronization metadata to the user. Consult
mirror.status only when data completeness affects the requested answer.
Async responses
Calls may return completed data or a job reference. Use --poll, or inspect a
job explicitly:
wannalead job -id <job_id>
wannalead cancel -id <job_id>
For invitations.send, a 202 status of
invitation_profile_resolve_pending or relationship_preflight_pending means
an internal prerequisite read is still running and no invitation job exists
yet. Poll the returned preflight_job_id, then replay the exact confirmed
request with the same idempotency key. Both wannalead invitations send --poll
and wannalead call invitations.send --poll perform this resume automatically.
The pending key is already bound to the canonical target and message; changing
either returns an idempotency conflict. If completed relationship evidence is
older than 60 seconds, the next exact replay refreshes it under the same key.
Internal prerequisite reads are unmetered; no product or LinkedIn account quota
is consumed until an invitation job is created.
Guarded writes
Write endpoints require linkedin:write, explicit confirmation, and an
idempotency key. Run a dry-run first when the command supports it. For generic
calls, inspect the endpoint's required headers and schema before sending.
Never infer missing write intent, retry a rejected send in a loop, or reuse one idempotency key for different operations.
A write job exists the moment the create call returns a job_id, even if your
process is interrupted, times out, or loses the --poll output afterwards. To
recover an interrupted write, replay the exact same request with the same
idempotency key (it returns the original job) or poll wannalead job -id <job_id>. Never resend under a new idempotency key: that creates a second job
and the recipient gets the message twice. A 409 idempotency_conflict answer
carries existing_job_id and poll_url when the key already owns a job; poll
that job instead of retrying.
For a standard campaign, prefer automation.campaigns.preview followed by
automation.campaigns.activate. The granular campaign endpoints remain the
supported custom builder for non-standard graphs.
Reporting a product issue
Use feedback.send when the user wants to report reproducible Wannalead
behavior such as an incorrect or incomplete result, an unclear contract, or a
workflow blocked by the product. Capture the endpoint name and any returned
request_id or job_id; state observed versus expected behavior. Never attach
credentials, cookies, raw provider payloads, or unrelated private content.
Feedback is an external write. Present the exact body and obtain the user's approval before sending it:
wannalead endpoints feedback.send --examples --json-schema
wannalead call feedback.send \
-body '{"message":"<observed behavior and expected behavior>","context":"endpoint=<name> request_id=<id> job_id=<id>"}'
The feedback endpoint is synchronous, separately rate-limited, and does not consume LinkedIn product quota. Do not retry-loop a rejected report.
For post writes, posts.comment accepts a post reference plus final text,
posts.comment-reply accepts parent_comment_urn plus final text, and
posts.repost accepts a post reference. Prefer activity_id/post_url; use
the URNs returned by posts.resolve only when chaining jobs. Wannalead owns the
canonical thread and root-content resolution.
Do not re-resolve messaging contacts by display name. Chain from
messages.conversations.participants[].profileAcoaa and reuse the exact
conversationUrn for thread reads or approved replies. Chain
events.attendees[].profileAcoaa directly into profiles.enrich.
For messages.conversations, continue from top-level data.next_cursor while
data.has_more is true. A full 20- or 25-row page is not terminal. Read the
first row's nextCursor only as a compatibility fallback for an older server.
Boundaries
Use only the public Wannalead catalog. Do not derive LinkedIn cookies, private identifiers, browser routing, gateway messages, or direct LinkedIn requests.
How can the creator link this skill?
Add the canonical catalog link to the repository README so users can inspect current installs and available audits. The publishing guide covers the complete discovery path.
<a href="https://skillzs.dev/skills/wannalead.co/endpoints">View endpoints on skillZs</a>