Skip to main content
Appeeky Spec ships six specialist subagents and a review that runs them, so your coding agent works like a small growth team. The specialists only read. The review decides, edits and writes the pull request. Nothing reaches the stores until a person merges it.

​
Install

Commit the files. Edit them to fit your app: running agents install again keeps your changes unless you pass --force. doctor checks that they are installed. In Claude Code, run /appeeky-cofounder. In agents without subagents, the review does each specialist’s checklist itself, one after another.

​
The specialists

Each specialist has read tools only (Read, Grep, Glob, Bash for the CLI) and answers with findings and up to two candidates. A specialist whose data is missing is skipped, and the brief says so.

​
The co-founder review

  1. Where things stand. Runs status --json; its free list is the stores with no release being measured. Reads appeeky/loops.md, appeeky/playbook.md and the last brief.
  2. Asks the specialists, in parallel when it can.
  3. Decides on the one change most likely to win, for a free store: one clear hypothesis, the strongest evidence, never a change the playbook says lost. If no store is free or no candidate is strong enough, it changes nothing and says why.
  4. Makes it. Edits appeeky/store/, runs check and diff, and fixes every error.
  5. Writes the brief to appeeky/briefs/<YYYY-MM-DD>.md: where things stand, what each specialist found, the change (or why there is none), what is for a person (review replies, ads changes, product requests, the next UGC experiment, posts to publish) and what it did not pick.
  6. Turns the social and UGC proposals into drafts in appeeky/social/drafts/ and appeeky/ugc/briefs/, checked with social check and ugc generate --dry-run, so they go out after the same merge. See Social posts and UGC clips.
  7. Writes .appeeky/proposal.md: a pull request title on the first line, then what changes, why, the metric and when it is checked.

​
Once a week

The weekly routine refreshes the data (data pull, reviews pull, competitors pull, social pull, ugc pull), finds the free stores, runs the review with Claude Code, and opens a pull request on a listing/weekly-<date> branch with the listing change, the brief, and any drafts and briefs. When every store has a release being measured, the pull request carries the brief only. The routine uses .github/appeeky-propose.md when you keep your own version of the review, and .claude/commands/appeeky-cofounder.md otherwise.

​
What the agent cannot run

The weekly routines start the agent with an allow list (file tools, subagents, appeeky-spec / npx @appeeky/spec, git log, git tag) and deny everything that publishes or spends: Dry runs stay allowed, so the brief can say what a clip or a scan would cost. After you merge, the push workflow or the Docker push routine sends what the pull request approved.