a field guide for agent context

ctxops

Agent skills for the work around the implementation loop: assemble the right context, pressure-test the direction, shape the output.

Agent Skills open standardMITv1.2.09 skillsmeasured, not vibes
/plugin marketplace add arniesaha/ctxops
/plugin install ctxops@ctxops

01 · Premise

The map is not the territory

An agent acts on the map you hand it (prompt, skills, context), not the territory of the real project. Where the map is silent, wrong, or stale is where agents fail, confidently. Most skill bundles optimize how the agent writes code. ctxops shrinks the map/territory gap at the bookends instead.

Before: pull the territory into the map. After: shape the output without losing the load-bearing facts.

02 · The set

Nine skills, two ends of the loop

ctx-pack

Assemble a tight working context pack for a task or decision.

project-briefing

One-page operational brief to resume or hand off a whole project.

recall-prior-work

Search persistent memory before answering "did we already do X?"

decision-review

Adversarial pass on a direction before committing. Post-hoc mode for old calls.

status-update

Stakeholder progress messages from git, memory, and open threads.

ticket-scoping

Turn a finding into paste-ready issues with checkable acceptance criteria.

artifact-packager

Turn real work into a durable shareable artifact with a publish checklist.

blindspot-pass

Interview + scan to surface a plan's unknowns before implementation starts.

ctx-ignore

Ranked noise report and agent ignore files. Exclusion is half of assembly.

Plain SKILL.md files per the Agent Skills standard: Claude Code, Codex, Gemini CLI, OpenCode, or any compatible harness. Trackers and memory tools are examples, not dependencies.

03 · Evidence

Measured, on real work

ctxops ships its own instrumentation and publishes its own numbers. From the first 26 days of instrumented maintainer use:

7/7
skills invoked, across 7 different projects (baseline before the trigger rework: one invocation, ever)
8·1·0
real invocations adopted · adapted · abandoned, classified from session transcripts
4-3-0
blind skill-on vs. skill-off evals: wins, ties, losses
EvalOnOffResultWhat decided it
ctx-pack86skill-onFreeform arm asserted a false load-bearing fact; the judge verified.
project-briefing87skill-onFreeform arm misreported repo push state.
recall-prior-work87skill-onSkill arm re-verified memory against the current repo, with dates.
ticket-scoping87skill-onSizes tied to driving assumptions; ACs tied to verification artifacts.
decision-review · status-update · artifact-packager88tiesCapable-model defaults already match these rubrics.

Every decided eval turned on the same failure: an unverified or stale load-bearing claim. That is the exact failure mode the skills' rules target. Method and caveats: evals/ab-results-2026-07.md.

04 · Feedback loop

It measures itself

  1. Measure: a hook logs every skill invocation, locally.
  2. Audit: skill_retro.py classifies each one as adopted, adapted, or unclear from your own transcripts.
  3. Feed back: structured issue templates and opt-in aggregate reports. Nothing sensitive leaves your machine.
  4. Change: new ideas incubate; they promote only on measured usage.
  5. Gate: ab_eval.py re-runs the blind eval before releases.
  6. Ship: semver bump, marketplace update.

05 · Install

Two commands

/plugin marketplace add arniesaha/ctxops
/plugin install ctxops@ctxops

Other harnesses: symlink or copy the skills/ directories into your skill path. A Codex-style manifest ships in the repo.