codevira AGENTS.md

Project instructions and persistent memory for Codevira, a tool that stores decisions and project context across coding sessions. They document fixed technical decisions and when to use Codevira's tools.

In plain words
What is it for?
Use them when working on a Codevira-tracked project, especially when handling storage, session data, startup behavior, hooks, or recorded engineering decisions.
Why use it?
They help coding agents follow existing project rules and avoid reversing decisions that were deliberately locked. They also direct agents to the approved ways to read and write project data.

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/sachinshelke/codevira/agents-md
Clone the repo
git clone --depth 1 https://github.com/sachinshelke/codevira

Made for: Codex, OpenCode.

Per session 2,638 This file is loaded in full into every session.
When invoked 2,638 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 $0.02638 $0.02638
Opus 5 $0.01319 $0.01319
Sonnet 5 $0.00528 $0.00528
Haiku 4.5 $0.00264 $0.00264

Measured 3d ago against content hash c427cba674b6, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

codevira 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 3d 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 · 122 lines

How it starts

The opening of the file, as written. The whole thing — 122 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Codevira-tracked project memory: codevira

Codevira — cross-IDE persistent memory. Read it with the codevira MCP tools (get_session_context, search_decisions); do not open .codevira/*.jsonl directly — those files are large and token-heavy.

Locked decisions (do_not_revert)

  • D000001 All disk writes in product surface MUST go through mcp_server/storage/atomic.py — never use open(..., 'w'), write_text(… · atomic, concurrency, storage, v3.0.0-rc-audit
  • D000002 All read paths for decisions / sessions / phases MUST go through the v3.0.0 JSONL canonical store (storage/decisions_st… · jsonl, read-path, storage, v3.0.0-rc-audit
  • D000006 analyze_session_outcomes() MUST run in a daemon thread at MCP server startup, never inline. The function spawns N git s… · mcp_server/server.py · mcp, outcome-tracker, perf, startup, v3.0
  • D000007 Hook script fast-path uses ~/.codevira/engine.disabled sentinel as a persistent alternative to CODEVIRA_ENGINE=0 env va… · mcp_server/data/hooks/user_prompt_submit.sh · cli, engine, hooks, perf, v3.0
  • D000008 repair_incomplete_init() uses a per-project .registered sentinel file to skip the global.db open once registration is c… · mcp_server/_repair_init.py · concurrency, global-db, perf, startup, v3.0
  • D000009 CODEVIRA_NO_WATCHER=1 env var skips start_background_watcher() in both stdio and HTTP MCP servers. Each codevira proces… · mcp_server/server.py · env-var, fsevents, perf, v3.0, watcher
  • D000010 Any change to a hero policy (mcp_server/engine/policies/*.py) — especially relevance_inject and decision_lock — MUST ru… · mcp_server/engine/policies/relevance_inject.py · e2e, hero-policy, regression, relevance-inject, testing, v3.0, wedge
  • D000012 The v3.0.0 JSONL store WRITE path now validates the resolved project root via is_invalid_project_root() inside storage/… · mcp_server/storage/paths.py · claude-desktop, forbidden-root, g5, ship-blocker, storage, v3.0.0, write-path
  • D00011Z [supersedes D00011W: D00011W claimed the data-loss bug was fixed and was locked do_not_revert on that basis. An adversa… · mcp_server/migrate.py · corrected, data-loss, high-severity, migration, release-blocker, root-cause, v3.7.1
  • D00012K [supersedes D0000PV: D0000PV's constraint is right and stands, but both of its numbers were retired by D00011J's measur… · budget, do-not-revert, footprint, invariant, lean, lightweight, performance
  • D00012X codevira clean is the UNINSTALLER, not a registry pruner — it removes hooks, unregisters from settings.json, and dele… · mcp_server/cli_uninstall.py · cli, destructive, gotcha, incident, naming
  • D00013X STANDING RULE (Sachin, 2026-08-16): everything in flight goes into 4.1.0, and NO other version is cut until 4.1.0 is pu…

Read the full file on GitHub · 122 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. 3d ago First seen · 122 lines · 2,638 tokens per session scan A c427cba674b6

Subscribe to this mod's changes

codevira AGENTS.md is an instructions file published in the GitHub repository sachinshelke/codevira (12 stars, last pushed 18d ago), licensed MIT. It adds 2,638 tokens to every session, about $0.0132 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.