opencode-loop-plugin AGENTS.md

A project instruction document for an OpenCode loop plugin. It explains the plugin's server, state storage, terminal UI, prompts, tests, and scheduling rules.

In plain words
What is it for?
Use it when modifying this plugin to understand where features live, preserve idle-session scheduling invariants, follow state-path and interval rules, and run the relevant Bun tests.
Why use it?
It gives coding agents the project-specific rules needed to change the plugin without breaking its loop lifecycle or scheduler behaviour.

Instructions file for CodexOpenCode

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add instructions/prevalentware/opencode-loop-plugin/agents-md
Clone the repo
git clone --depth 1 https://github.com/prevalentWare/opencode-loop-plugin

Made for: Codex, OpenCode.

Per session 982 This file is loaded in full into every session.
When invoked 982 The same file — it is already loaded in full.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00982 $0.00982
Opus 5 $0.00491 $0.00491
Sonnet 5 $0.00196 $0.00196
Haiku 4.5 $0.00098 $0.00098

Measured yesterday against content hash cdb0c9a53239, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

opencode-loop-plugin AGENTS.md scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured yesterday.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

AGENTS.md · 64 lines

How it starts

The opening of the file, as written. The whole thing — 64 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Agent Notes

Project Shape

This package is an OpenCode plugin with separate server and TUI entrypoints:

  • src/server.ts is the server plugin. It registers the /loop command, the loop tools, chat/session hooks, compaction context, and the scheduler that injects loop iterations while a session is idle.
  • src/state.ts owns loop persistence and lifecycle state. It stores JSON at OPENCODE_LOOP_STATE_PATH when set, otherwise under the user's OpenCode data location. It also owns interval parsing/formatting.
  • src/tui.tsx is the Solid/OpenTUI sidebar and command-palette UI. It is exported as source, so avoid adding heavy runtime dependencies here unless there is a strong reason.
  • src/prompts.ts contains the /loop command template, the per-iteration synthetic prompt, the loop system reminder, and the compaction context.
  • test/ covers state, server hooks/tools/scheduler, and TUI helpers with Bun tests. Scheduler tests use 1-second intervals via the min_interval_seconds option.

Scheduler Invariants

  • Iterations are only injected while the session is idle; busy sessions defer with min(intervalMs, busy_backoff_seconds).
  • sendingLoops guards against concurrent iterations of the same loop; after a successful injection the session is optimistically marked busy so sibling loops defer.
  • Dynamic loops live one iteration at a time: dynamicPending entries settle on idle, and a dynamic loop whose turn ended without schedule_next_run or stop_loop is stopped. sawBusy must start false for injected iterations (only the creating turn starts true) so a stale idle event cannot settle a just-injected loop.
  • runDue must never throw: it runs from setTimeout callbacks, so errors are caught and logged.
  • Timers are per-loop, always re-armed from persisted nextRunAt, and cancelled on stop/pause/dispose.

Change Guidelines

  • Preserve the public Promise-based state API (createLoop, listLoops, stopLoop, recordRunSent, etc.) because OpenCode hooks and tests call it directly.
  • Keep zod pinned to the exact version @opencode-ai/plugin depends on; tool argument schemas fail typechecking across zod minor versions.
  • Effect is intentionally used in the state/persistence boundary. Do not spread Effect into the TUI.
  • State writes should remain atomic: write to a temp file, then rename into place.
  • Use OPENCODE_LOOP_STATE_PATH for tests and smoke runs so you do not touch a real user's loop state.
  • All loop tools return { ..., loops } for the session; the TUI sidebar parses that shape from tool outputs, so keep it stable.

Read the full file on GitHub · 64 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. yesterday First seen · 64 lines · 982 tokens per session scan A cdb0c9a53239

Subscribe to this mod's changes

opencode-loop-plugin AGENTS.md is an instructions file published in the GitHub repository prevalentWare/opencode-loop-plugin (3 stars, last pushed 3d ago), licensed MIT. It adds 982 tokens to every session, about $0.0049 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.

Related

Other instructions, from other repositories