Approving and rejecting tactics

Approval is the moment the platform actually does something. Knowing exactly what each click triggers, and when to skip, is what separates a careful operator from one who just rubber-stamps.

What happens when you approve

When you click Approve and start, the proposal status flips to approved and Outrigger starts the matching work. The routing is simple:

• review_velocity, review_platform_breadth → review/scan
• directory_claiming, linkedin_company_page, wikipedia_eligibility → entity/scan
• reddit_seeding, quora_seeding → discovery/scan
• press_distribution, youtube_* → no auto-route: you drive these manually

The same Inngest jobs that power the rest of the platform run with the agent's payload as input. Nothing new, nothing experimental: the agent just stages the work for the engines you already trust.

Where client action records are enabled, approval also saves the work in the action ledger as dispatched. From there your team can report implementation with evidence. Approval itself never proves the work went live or changed visibility.

If a proposal files work from another area under the wrong tactic (for example on-page work as directory claiming), approval is refused with a message pointing to the right workspace. Reject it and handle the task there.

When to reject

Reject when:

• The client isn't actually ready for this tactic yet (entity not canonical, brand brief still shaky, no audit baseline).
• The payload targets the wrong audience (gap clusters look off-topic, subreddits are unrelated, directories aren't relevant for the vertical).
• A teammate already handled this manually in the last few days.

Rejecting tells your team the proposal was not right for this client. It does not retrain the agent, so add a short reason that a teammate can learn from.

When to override manually

Sometimes you know the right move is something the agent didn't propose: a press push tied to a launch, a review campaign before a renewal, a custom citation seed. Run those tactics manually from the relevant module. The agent sees that work as already in progress and will not propose a duplicate.