Update to Compound Engineering

{
  "description": "Retune a skill corpus for a new model, measurement-first: mine the run archive for a baseline, establish a noise floor, audit the corpus adversarially, then cut in measured passes until a pre-registered bar clears. Requires a benchmark harness that can A/B two builds of the corpus; refuses without one.",
  "included_files": [
    {
      "relative_path": "references/baseline-mining.md",
      "size_in_bytes": 10649
    },
    {
      "relative_path": "references/corpus-audit.md",
      "size_in_bytes": 12813
    },
    {
      "relative_path": "references/cut-passes.md",
      "size_in_bytes": 10877
    },
    {
      "relative_path": "references/halt-taxonomy.md",
      "size_in_bytes": 11221
    },
    {
      "relative_path": "references/noise-floor.md",
      "size_in_bytes": 12130
    },
    {
      "relative_path": "references/workflow-shapes.md",
      "size_in_bytes": 11500
    }
  ],
  "name": "ce-retune",
  "skill_md_contents": "---\nname: ce-retune\ndescription: \"Retune a skill corpus for a new model, measurement-first: mine the run archive for a baseline, establish a noise floor, audit the corpus adversarially, then cut in measured passes until a pre-registered bar clears. Requires a benchmark harness that can A/B two builds of the corpus; refuses without one.\"\ndisable-model-invocation: true\nargument-hint: \"[target model or symptom] [path to the corpus, defaults to ./skills] [bar:<n> consecutive clean runs]\"\n---\n\n# Retune a Corpus for a New Model\n\nA corpus that degrades on a new model is a measurement problem before it is a writing problem: rewriting what looks wrong produces a plausible fix list and no way to know whether any item mattered.\n\n**Outcome:** a corpus whose measured behavior on the target model clears a bar registered before any change, with the regression classes removed and each removal attributable.\n\n**Done:** the bar is cleared, or the run reports the specific claim it could not support. A green test suite is not done: it proves nothing broke, not that behavior improved.\n\n**Non-goal:** word reduction. Leanness and performance are separate programs that share a corpus; only one of them is the result here. Report completion, not word count.\n\n\n## Phase 0: the measurement gate — check this first\n\nThis skill cannot run without a way to observe behavior. Check for all three, and name whichever is missing:\n\n1. **A run archive or a harness that produces one** — per-run logs carrying the tool-call trace, a terminal marker, token counts, and the final message.\n2. **A build selector** — the harness can point a run at a specific source checkout of the corpus (a `--plugin-dir`-style override, a configurable skills path, an env var), so two builds are comparable under one runner.\n3. **A repeatable task** the corpus actually executes end to end.\n\nIf any is missing, **stop and say so**, naming what to build. Do not fall back to a static audit and present it as retuning: an audit can say what looks cuttable and never whether cutting helped. An audit-only pass is a legitimate thing to want; it is a different request.\n\nState the target model and the harness you found before continuing.\n\n## The phases\n\nThey run in order, and each names the reference it cannot start without. Read `references/workflow-shapes.md` before dispatching any phase: the wrong orchestration shape is the common failure. Fan out by disjoint file ownership, never by item. Items cross files, and agents that share a file lose each other's edits.\n\n1. **Mine the archive** before spending a run — `references/baseline-mining.md`. Historical runs are a free baseline, usually larger than any experiment affordable now.\n2. **Establish the noise floor** — `references/noise-floor.md`. Run the harness against **two identical copies** of the corpus, same commit on both sides; whatever difference appears is the floor every later claim must clear. **Register the bar now, in writing, before any change exists.** A bar chosen after seeing results is not a bar.\n3. **Audit the corpus adversarially** — `references/corpus-audit.md`. One agent per skill proposes cuts; a second per skill does the opposite and defends the existing prose. **The two passes require independent contexts.** If the host exposes no way to run them as separate agents, report that as a blocker and stop the audit — do not argue both sides in one context and present the result as an audit.\n4. **Cut in surgical passes**, one problem per agent — `references/cut-passes.md`, and `references/halt-taxonomy.md` when the symptom is stalling, halting, or a run that ends while naming work it did not do. Two rules bound every pass, whatever class it is cutting. **Never edit a test to make a suite green**: a removed string a test pins is a finding to report, not a test to weaken. And **not every stop is the enemy.** Some workflows exist to stop and ask; that is the product. Sort every stop by who is actually on the other side before touching it. `references/halt-taxonomy.md` carries the screens that decide, so read them before cutting any stop.\n5. **Measure, then let the failure choose the next fix** — `references/cut-passes.md` again for what each failure site means and for auditing the phases the instrument never enters. Loop 4 and 5 until the registered bar clears. Then stop; a bar cleared is done. Also report what stayed unmeasured: a cleared bar never implies coverage it does not have.\n6. **Ship** — `references/cut-passes.md` carries what the commits and the write-up must preserve.\n"
}