ads

Plugin ads from marketplace ads-discipline in tka4enko/ads-discipline. Commands read from the repo's manifest — 1 month ago.

Skills (7)

• daily

  The daily routine - snapshot what nothing can recover, diff against yesterday, record detected changes with baselines, raise decisions due for verification, write the brief. Run by a scheduled task, or by hand with "run the daily routine". Not interactive - it asks nobody anything; what it cannot an

• explain

  Explain an advertising term, metric, status or account behaviour to someone without a marketing background — using their own data. Use for "what does this mean", "why is it like this", "what is X", "is this normal or not".

• export

  Turn live connector reads into the portable CSV export that an audit can consume. Use before running any paid-media audit that needs account data, when an audit reports zero evidence or insufficient coverage, or for "prepare the data for the audit", "export the account", "why does the audit say it h

• gaps

  Registry of data gaps — what is missing to reach a conclusion, where to get it, and what it would unlock. Use when asked "why can't you tell", "what data is missing", "what should we connect", or whenever another skill hits a missing source.

• orient

  Establish what the account can and cannot currently be trusted to tell you — which sources are connected, which conversion counters are still firing, which numbers are unusable and why. Use when asked "what's going on", "where are we", "what do we actually see", "is anything broken", or before trust

• setup

  Check what this session is missing and print exactly how to fix it — which connectors respond, whether a state repository is named and reachable, and what each missing piece would unblock. Run this first on a new machine, a new surface, or when an answer says it cannot see something. Use for "what d

• sharpen

  Interrogate a campaign, goal, ICP or hypothesis until it is sharp enough to test — grounded in what the account actually says. Use for "who is this campaign even for", "what counts as a result here", before launching anything, setting a target, or whenever a statement sounds vague ("let's try a new

Subagents (4)

• analyst

  Main-thread persona for advertising work. Applies the confidence-label and missing-data discipline to every answer, not only inside a skill. Activated automatically while this plugin is enabled.

• collector

  Pulls raw numbers from ad platforms, GA4 and the CRM. Does not interpret or evaluate — only retrieves and normalises. Invoke for any data gathering that precedes analysis.

• diagnoser

  Works out why an object behaves the way it does — underdelivery, a jump in cost, a drop in conversion. Connects metrics, delivery statuses, change history and CRM outcomes. Invoke after collector has gathered the numbers and a deviation has been found.

• judge

  Rules on contested verdicts and synthesises the playbook — decides whether an accumulated set of experiments justifies promoting a rule, and whether existing rules still hold. Invoke rarely, for decisions whose cost of error is measured in months.

MCP servers (3)

• meta-ads

  connects to: https://mcp.facebook.com/ads

• hubspot

  connects to: https://mcp.hubspot.com/anthropic

• github

  connects to: https://api.githubcopilot.com/mcp/

Hooks (1)

• SessionStart

  "${CLAUDE_PLUGIN_ROOT}"/hooks/pending-context.sh

Similar plugins