Other agents

opencode

opencode reads both ~/.agents/skills and ~/.claude/skills, so slopcop init covers it too. Ask it to check its change, or add the usual line to your project’s AGENTS.md:

Run slopcop before you finish and fix what it reports.

There is no Stop hook for opencode; the skill is how it knows to check.

Any agent with a shell

Any agent that can run a shell command can run slopcop: it never prompts, sends one request, and answers with an exit code. Run outside Claude Code or Codex, a check sends nothing about the session, only the repository URL, the base commit, and the patch.

If the agent cannot load skills, give it the essentials in its instructions file:

Before you finish, run `slopcop --format json` from the repository root and
read the report.
- Exit 0: clean.
- Exit 1: the change added findings. Fix each one and run the check again.
  Read a finding whose certainty is "possible" or "heuristic" before changing
  code; if it is a false positive, run `slopcop ignore <id>` with its id.
- Exit 2: some rule could not finish, or the check failed. Say which; never
  treat it as clean.
- Exit 3: still running. Run the same command again.
- Exit 4: refused; only a person can clear it. Stop and report the code and
  remedy from the output.
When checking committed work, pass `--base <branch the change targets>`.

The full instructions are in the skill itself, which you can read at skills/slopcop/SKILL.md and adapt.

Signing in for an agent

An agent runs with your login. On a machine where you cannot open a browser, store a token instead of running the device flow:

slopcop login --token "$SLOPCOP_TOKEN"

or set SLOPCOP_TOKEN in the agent’s environment, which needs no login at all. See Sign in. Give agents that share a login their own --workspace rather than switching the login’s workspace; see Workspaces.