contexture codecs.instructions.md

Coding rules for working with hardware video codecs, which compress and decompress video using systems such as NVENC, VideoToolbox, or MediaCodec.

In plain words
What is it for?
Managing encoder and decoder lifetimes, keeping one session for a stream, reconfiguring it when supported, and draining or flushing data before destruction.
Why use it?
They help avoid leaked hardware sessions, unnecessary per-frame setup, broken timing, and incomplete output during shutdown.

Instructions file for GitHub Copilot

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/ackeskin/contexture/codecs
Clone the repo
git clone --depth 1 https://github.com/AcKeskin/contexture

Made for: GitHub Copilot.

Per session 1,625 This file is loaded in full into every session.
When invoked 1,625 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 $0.01625 $0.01625
Opus 5 $0.00813 $0.00813
Sonnet 5 $0.00325 $0.00325
Haiku 4.5 $0.00162 $0.00162

Measured yesterday against content hash 68466c2729bb, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

contexture codecs.instructions.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.

.github/instructions/codecs.instructions.md · 44 lines

How it starts

The opening of the file, as written. The whole thing — 44 lines — stays where its author put it; the contents beside it link to each section on GitHub.

codecs rules

Auto-loaded by Copilot when editing files matching **. Generated from architectural-rules/codecs/ — do not hand-edit.

Codec session lifetime

Treat each hardware encode/decode session as an explicitly-owned scarce resource with a defined create→use→destroy lifetime; release it deterministically, never on GC/finalizer timing. Hardware codec engines (NVENC/NVDEC, VideoToolbox, MediaCodec) expose a small fixed number of concurrent sessions; a leaked session starves every other consumer on the host until process exit. (VideoToolbox VTCompressionSession / VTDecompressionSession; NVENC nvEncOpenEncodeSessionEx + nvEncDestroyEncoder contract)

Open the session once for a stream and feed every frame through it; never create-and-destroy per frame. Session creation negotiates hardware, allocates the DPB/reference pool, and parses/builds parameter sets — per-frame churn destroys rate-control state, forces an IDR each time, and dominates latency. Reconfigure in place where the API allows (bitrate/framerate change) rather than tearing down. (NVENC reconfigure; VTSessionSetProperty; H.264/AVC §7.4.2 parameter-set continuity)

FLUSH/drain the codec before destroying the session and only then release. Encoders and decoders hold queued/in-flight frames (lookahead, B-frame reordering, async slots); destroying without draining silently drops trailing frames and truncates the bitstream. Submit the end-of-stream/flush sentinel, pump completions until the codec reports empty, then destroy. (VTCompressionSessionCompleteFrames; NVENC EOS NULL-input flush; MediaCodec end-of-stream BUFFER_FLAG_END_OF_STREAM)

For callback/async-completion codecs, treat submit and completion as separate events: do not assume output is ready when input returns, keep input frames/buffers alive until their completion fires, and join all outstanding completions before teardown. Releasing an input buffer or the session while a completion is pending corrupts output or faults the driver. (VTCompressionSession output handler; NVENC async event + nvEncLockBitstream; MediaCodec dequeueOutputBuffer)

Read the full file on GitHub · 44 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 · 44 lines · 1,625 tokens per session scan A 68466c2729bb

Subscribe to this mod's changes

contexture codecs.instructions.md is an instructions file published in the GitHub repository AcKeskin/contexture (2 stars, last pushed 1mo ago), licensed MIT. It adds 1,625 tokens to every session, about $0.0081 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.