gigayaya

Context Render — Development skill for Claude Code

Development community

Context observability for Claude Code: see which of your scaffolding components an agent actually used.

How to install Context Render

This entry records only its repository, not the path inside it, so there is no exact command to give. Open gigayaya/context-render and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Context Render does

Context observability for Claude Code: see which of your scaffolding components an agent actually used.

Alternatives in Development

  • Cost Xray — See what Claude Code and Codex actually send to the API — and what each part costs 586 ★
  • New Project — Create a new project with all scaffolding rules applied 335 ★
  • Claude Code Otel — A comprehensive observability solution for monitoring Claude Code usage, performance, and costs 306 ★

README

context-render

Which of your Claude Code scaffolds — skills, commands, subagents, MCP servers, hooks, CLAUDE.md files — actually get used?

`context-render` reads your local session transcripts (read-only) and reports each component's state per session: **R** registered → **L** loaded → **I** invoked. No scores, no API calls, no telemetry — everything stays on your machine.

Install

Requires Python 3.11+. Install straight from GitHub:

pip install git+https://github.com/gigayaya/context-render.git

Or clone first:

git clone https://github.com/gigayaya/context-render.git
cd context-render
pip install .

Either way you get the `ctxr` command on your PATH.

Quickstart

cd your-repo
ctxr init           # scan scaffolds → .context-render/manifest.yaml
ctxr sync           # ingest past sessions (idempotent, safe to re-run)
ctxr sessions       # list ingested sessions, newest on top
ctxr sessions   # one session: what was loaded/invoked, full timeline

Once you have some history:

ctxr report --since 30d      # cross-session aggregate: active / low-use / unused, plus what the agent had to go find itself
ctxr map                     # static: is your guidance a usable routing map, and what does it fail to cover?

What a session report looks like

Every session report (`sessions `) has three views of the same session:

**File loads** — every file that entered the context, in injection order, with how it got there (`Read`, `Bash`, system-injected) and, where possible, which component it's attributed to:

![File loads: context injection order with load mechanism and attribution](docs/images/file_load.png)

**Timeline** — the session as a chronological event list: hooks firing, CLAUDE.md injection, reads, bash commands, writes. `[L]` marks a load, `[I]` an invocation; `~` marks heuristic (vs. exact) attribution:

![Timeline: chronological session events with L/I state markers](docs/images/timeline.png)