- Codex adapter (
packages/codex-plugin, hint-only: nothing outside a Codex session can run/compact). Its sources are TypeScript that plainnodetype-strips, so they must stay erasable - no parameter properties, no enums. Codex rebuilds a hook's PATH from a shell snapshot and may have nonodeon it, so hooks go throughhooks/run.sh, which is silent when it finds none. Judging is gated to the interactive TUI by the rollout'soriginator. The rollout transcript format is explicitly unstable upstream:src/rollout.tsnever throws and degrades to the strictest floor.scripts/validate.mjschecks the package is loadable exactly as shipped. Snooze/dismiss are a known gap and follow-up: suppressing advice safely needs a reliable current-session identity that Codex gives neither the hook nor the CLI, so the CLI could only mutate the most recently written session record. Windows support is a possible follow-up: this ship is macOS and Linux only because the hooks run through POSIXshand have nocommandWindowsentries.
/codex-adapter
@37fa1a6Use when modifying or testing `packages/codex-plugin` or Codex `/compact` behavior.
- 1 file
- 1.2 KB
- Updated 2 months ago
- GitHub
Use this Skill: https://skilld.dev/gh/kunchenguid/compact-adviser/codex-adapter Nothing lands on disk. Nothing to clean up.
Edit a local copy. It keeps the author and licence.
≈25 tokens for metadata: the name and description. ≈261 when used: this file.
Description uses 1.2% of example budget
Before choosing a Skill, your Agent reads its name and description. All available Skills share that space.
- A shorter description leaves more room for other Skills. This entry exceeds our 1% size suggestion.
In our Claude Code example, all Skill names and descriptions share 8,000 characters. This Skill uses ≈99 characters, or 1.2%.
The 1% threshold is a size suggestion. Longer descriptions can still fit.
Your model, settings, and other Skills decide how much text your Agent can read.
Example settings and source
The example uses a 200k-token context and default Claude Code settings. The count includes the name, description, separators, and when_to_use when present. Codex also counts local file paths.
Skit's source and limits: Codex 0.160.1, Claude Code 2.1.292.
Source: SKILL.md on GitHub
Skill behaviors
No rule matched.
skilld matched fixed text patterns in SKILL.md and file names. Patterns miss obfuscated code.
skilld run checks every file with the same patterns. It asks for approval before it loads a Skill with a behavior marked Needs approval.
Third-party checks
Provenance
Signed by skilld at 37fa1a6. This ties the file your Agent reads to that commit on GitHub. It does not review the instructions.
Last checked against GitHub 4 weeks ago.
Capability
- user-invocable
- false
- metadata
{ "internal": true }