Skip to content

Reposition README: local black box recorder for Codex CLI - #1

Merged
kkrlstrm merged 1 commit into
mainfrom
reposition-readme
Jul 6, 2026
Merged

kkrlstrm merged 1 commit into
mainfrom
reposition-readme

Conversation

@kkrlstrm

@kkrlstrm kkrlstrm commented Jul 6, 2026

Copy link
Copy Markdown
Owner

Repositions the repo around its actual insight instead of a generic "telemetry" framing.

Why

The old headline read "Telemetry for the OpenAI Codex CLI — the sibling of cc-logger." Accurate but generic, and it buried the sharpest point: Codex hooks are incomplete (on the build tested, PreToolUse/PostToolUse fire for shell only), so this tool bypasses hooks and reconstructs the full run — apply_patch, MCP calls, subagents — from the rollout files Codex already writes. That's the wedge, and it now leads.

What changed

  • New H1 + tagline — "The local black box recorder for OpenAI Codex CLI," full-fidelity local observability.
  • Pain-first structure — "Why this exists: hooks are not enough" comes before the cc-logger sibling relationship.
  • Real proof — actual codex-logger stats output up top (apply_patch / write_stdin counts that hooks would miss), plus sessions/inspect examples.
  • "Questions it helps answer" section to make the value concrete.
  • Copy-paste SQL for the three most common questions (top cost, tool failure rate, subagent share).
  • Capability comparison table (hooks vs cloud tracing plugin vs codex-logger) with a caveat that the hooks row is version-specific.
  • "Where this fits" — the local agent-ops stack (cc-logger / agent-guard / codex-logger).
  • Softened the version-pinned hook claim so a future Codex release that changes hook coverage can't make the headline read as wrong.
  • Updated the pyproject.toml package description to match.

Docs only — no code changes. Stats/sessions output in the README is real (ingested from local Codex history); the inspect block is sanitized/illustrative to avoid leaking real paths.

🤖 Generated with Claude Code

Lead with the actual wedge (hooks miss apply_patch/MCP/subagents, so
reconstruct the full run from rollout files) instead of a generic
"telemetry" framing. Pain-first structure, real stats output as proof,
"questions it answers" + copy-paste SQL, an honest capability table,
and the agent-ops-stack context. Softens the version-pinned hook claim
so a future Codex release can't make the headline read as wrong.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
@kkrlstrm
kkrlstrm merged commit 3a308f9 into main Jul 6, 2026
4 checks passed
@kkrlstrm
kkrlstrm deleted the reposition-readme branch July 6, 2026 14:42
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant