context-management

A set of practices for protecting large data during long coding-agent sessions. It covers files such as screenshots, videos, base64 data, JSON, and lengthy test output.

In plain words
What is it for?
It helps pass large artifacts between turns, manage long test runs, and keep session data reliable.
Why use it?
It prevents context limits or compression from silently changing data and producing corrupt uploads or incomplete results.

Skill for Claude CodeCodex

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 skills/implicit-labs/autosymph/context-management
Any agent
npx skills add implicit-labs/autosymph --skill context-management
Clone the repo
git clone --depth 1 https://github.com/implicit-labs/autosymph

Made for: Claude Code, Codex.

Per session 56 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 2,005 The whole file, excluding the scripts and references it only reads on demand.
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 $0.00056 $0.02005
Opus 5 $0.00028 $0.01002
Sonnet 5 $0.00011 $0.00401
Haiku 4.5 $0.00006 $0.00200

Measured 2d ago against content hash 1ba52d9b4700, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

context-management 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.

The scan reads SKILL.md. This mod also ships 1 executable file (test.sh), listed below but not scanned — reading those needs a real analyzer, not pattern matching.

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.

skills/context-management/SKILL.md · 199 lines

How it starts

The opening of the file, as written. The whole thing — 199 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Context Management for Long-Running Agents

Prevent context compaction from silently corrupting your data. This skill teaches three layers of defense: file-based handoff, size awareness, and context threshold management.

When to Use

  • During verify runs that capture multiple screenshots or videos
  • When base64-encoding files for Linear upload
  • When test output exceeds a few hundred lines
  • When a session is running long (35+ turns)
  • When tool results start showing "persisted to disk" messages
  • Any time you hold large data in context across API turns

Layer 1: File-Based Handoff

The core rule: Never hold large data in context across API turns. Write it to disk, then read it back in the SAME turn you need it.

Context compaction triggers at ~80% window fill. Any data in older turns is fair game for truncation — silently. You won't get a warning. Your 18,984-char base64 becomes 8,498 chars and the upload produces a corrupt file.

Pattern: Write → Read-in-Same-Turn → Consume

# Step 1: Generate data and write to disk (ONE Bash call)
sips -Z 800 -s format jpeg input.png --out output.jpg 2>/dev/null && \
  base64 -i output.jpg > output.b64 && \
  wc -c < output.b64

# Step 2: Read file back (in the SAME response as Step 3)
cat output.b64

# Step 3: Consume immediately (SAME API turn as Step 2)
mcp__linear__create_attachment(
  issue: "ISSUE-XXX",
  base64Content: <content from cat>,
  filename: "output.jpg",
  contentType: "image/jpeg",
  title: "Description"
)

Steps 2 and 3 MUST be in the SAME response. If they're in different API turns, context compaction may truncate the base64 between them.

When to Use Files vs Inline

Data Type Size Action
Base64 (any) >10K chars Write to .b64 file
Base64 (any) <10K chars OK inline if consumed same turn
JSON payload >50 lines Write to .json file
Test output >200 lines Pipe to file: swift test 2>&1 | tee test-output.txt
Screenshot PNG Any Always resize first (see Size Thresholds)
Video Any Always file-based (too large for context)
Git diff >100 lines Write to file, read relevant sections

Read the full file on GitHub · 199 lines

Files

What ships with it

1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.

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 · 199 lines · 56 tokens per session scan A 1ba52d9b4700

Subscribe to this mod's changes

context-management is a skill published in the GitHub repository implicit-labs/autosymph (5 stars, last pushed 8d ago), licensed MIT. It adds 56 tokens to every session and 2,005 once invoked, about $0.0003 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 skills, from other repositories

systematic-debugging

Use when encountering any bug, test failure, or unexpected behavior, before proposing fixes.

obra/superpowers · 21 tokens

brainstorming

You MUST use this before any creative work - creating features, building components, adding functionality, or modifying behavior. Explores user intent, requirements and design before implementation.

obra/superpowers · 37 tokens

chat-pet-sprite-creation

Use when creating or changing VS Code chat pet sprite art, sprite sheets, state animations, eye treatments, Stable/Insiders variants, or pet transitions under src/vs/workbench/contrib/chat/browser/widget/media/chatPet.

microsoft/vscode · 53 tokens

cpu-profile-analysis

Analyze V8/Chrome CPU profiles (.cpuprofile) and DevTools trace files (Trace-.json). Use when: profiling performance, investigating slow functions, comparing code paths, finding bottlenecks, analyzing timeToRequest, understanding call trees from sampling profiler data, analyzing layout/paint/rendering, investigating…

microsoft/vscode · 71 tokens

agent-host-chat-contributions

Build and review cross-cutting agent-host chat behavior through lifecycle contributions. Use when adding turn lifecycle side effects, prompt or context injection, restored-history transformation, protocol-action observation, or when reviewing changes that add code to AgentSideEffects or AgentService.

microsoft/vscode · 56 tokens

auto-perf-optimize

Run agent-driven VS Code performance or memory investigations. Use when asked to launch Code OSS, automate a VS Code scenario, run the Chat memory smoke runner, capture renderer heap snapshots, take workflow screenshots, compare run summaries, or drive a repeatable scenario before heap-snapshot analysis.

microsoft/vscode · 62 tokens