codex-usage-tracker AGENTS.md

Repository instructions define how to work on a local Codex usage tracker. They explain which documents control the project and how separate work items are assigned and handed off.

In plain words
What is it for?
Use them when implementing tracker features, following the project roadmap, checking task ownership, or handing work between agents.
Why use it?
They help agents avoid relying on outdated plans and keep facts, calculations, ownership, and task progress consistent.

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/douglasmonsky/codex-usage-tracker/agents-md
Clone the repo
git clone --depth 1 https://github.com/douglasmonsky/codex-usage-tracker

Made for: Codex, OpenCode.

Per session 5,241 This file is loaded in full into every session.
When invoked 5,241 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 $0.05241 $0.05241
Opus 5 $0.02620 $0.02620
Sonnet 5 $0.01048 $0.01048
Haiku 4.5 $0.00524 $0.00524

Measured 2d ago against content hash 302e3f9c9d3f, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

codex-usage-tracker 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 2d ago.

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 · 420 lines

How it starts

The opening of the file, as written. The whole thing — 420 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Codex Usage Tracker Instructions

Product and authority

This repository is replacing the released 0.28 spike with a clean, local, Codex-first workflow-observability kernel. The kernel owns exact facts, identity, ordering, deduplication, calculations, freshness, coverage, allowance, valuation, bounded queries, and stable evidence. The consuming model owns interpretation, hypotheses, prioritization, caveats, and recommendations.

Start at docs/INDEX.md. The only implementation roadmap is docs/roadmap/AGENT_FIRST_CLEAN_CUTOVER.md; its checkbox ledger is docs/roadmap/TASK_PACKETS.md. The remaining dependency graph, delegation readiness, ownership locks, and allowed parallelism are controlled by docs/roadmap/REMAINING_EXECUTION_PLAN.md. Each delegable unit has one file under docs/roadmap/tasks/. Archived documents and historical release notes are evidence, not authority.

CK-07C additionally owns docs/architecture/PLAN_OPERAND_AND_FACT_CONTRACT.md and the versioned config/agent-kernel/plan-operand-contract-v1.json. CK-07D merged the effective-dated rate-card valuation correction at e49531b. CK-07E is the admitted test-only prerequisite for independent structural and query-only database-v1 fact adapters. CK-07A consumed CK-07B through CK-07E and is complete with 80 / 80 structural-v2 variants requalified. CK-08 is complete on merge with all 21 plans and 42 variants qualified as fact-backed mechanism evidence. A downstream architecture audit found that the expected-answer lane shares production evaluation, runtime keyset pagination follows complete Python materialization, projection classification combines execution stages, and publication/evidence scale plus replacement maintainability need corrective proof. CK-08R0 froze corrective-gates-v1; CK-08R2 is complete and CK-09 remains blocked. CK-08R1A froze corrected answer meaning and recursive closure; R1C is accepted at exact main fb0c578, and R1B is accepted at exact main 9e9332b3. Final R1 requalification passed hosted CI in PR #439, squash-merged, and was exact-main verified at 0832b854. CK-07R1's separate post-terminal roadmap completion makes CK-08R4 the sole Ready packet while CK-08RG and CK-09 remain blocked. CK-QG1A removed the two R2 page-executor complexity findings without changing behavior or the frozen baseline and is accepted at exact main 30983d4; QG1 PR #392 passed hosted CI, squash-merged, and was exact-main verified at 68050b93. CK-07R1A corrected the exact hosted Python 3.14 lifecycle-tail blocker; the linked CK-07R1A0 authorities, including argv correction, are merged through 479cbdb. Coordinator disposition and clean exact-main reapplication from 6c08ecd9 derived the exact 66c015de… / f108dbb4… / 4c514889… candidate cohort. The versioned shared-successor-overlay-authority-v1 preserves accepted CK-08R1B, CK-08R1, and CK-QG1 bytes while admitting only that complete cohort as CK-07 worker_prequalification. The sole CK-07R1 v2 launch is terminal failed_after_launch; its token is consumed and non-refundable, no planner-valid receipt or output exists, and no retry, restart, replacement, or further invocation is permitted. PR #448 merged the exact deterministic planner-selected small/large correction and immutable v1/v2 terminal evidence at exact main 1d0466b1. The additive lifecycle-post-terminal-completion-authority-v1 accepts that merged deterministic evidence only for CK-07R1 roadmap dependency completion. It permanently records runtime_acceptance=not_claimed, planner-valid receipt absent, post_single_run unavailable, and final_accepted unavailable. It neither reclassifies the failed run nor changes production semantics. CK-08R4 must independently measure current merged publication behavior and must not claim the missing CK-07R1 runtime acceptance. PR #394 remains stale failed read-only. Retained R3 evidence proved the EvidenceService outer query physically unbounded; CK-08R3A owns that isolated fix and R3 awaits its accepted, merged, exact-main-verified result.

Read the full file on GitHub · 420 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. 2d ago First seen · 420 lines · 5,241 tokens per session scan A 302e3f9c9d3f

Subscribe to this mod's changes

codex-usage-tracker AGENTS.md is an instructions file published in the GitHub repository douglasmonsky/codex-usage-tracker (193 stars, last pushed 12d ago), licensed MIT. It adds 5,241 tokens to every session, about $0.0262 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 2026-08-30.