AGENT SKILL

apt_recovery_evaluation_harness

Build bounded APT recovery experiments with executable evidence, generated-skill invocation logs, source manifests, and mechanism checks.

License 许可
Apache-2.0
最近更新
2026-09-03
Stars
★ 351
Forks
27

01 前言 Front Matter

name
apt_recovery_evaluation_harness
description
Build bounded APT recovery experiments with executable evidence, generated-skill invocation logs, source manifests, and mechanism checks.

02 资源文件 2

03 SKILL.md 正文

APT Recovery Evaluation Harness

Use this skill when a Distiller recovery run needs to test whether generated APT skills can recover the paper's mechanism without reading the original implementation repository. The harness is especially useful in soft mode, where a declared reduced/proxy experiment may be accepted only after executable evidence and validation pass.

Inputs

  • Attempt directory with module_plan.json, paper_profile.md, and environment/runtime_handoff.json.
  • Generated skills root containing the APT protocol, transformation, and atomic-loss skills.
  • Recovery mode from the normalized run config.
  • Bounded command budget and source-boundary constraints.

Outputs

  • recovery/experiment_plan.md.
  • recovery/source_manifest.json.
  • recovery/logs/experiment_command_log.json.
  • recovery/logs/generated_skill_invocations.json.
  • recovery/logs/generated_data_item.json.
  • recovery/logs/training_trace.json.
  • recovery/recovery_result.json.

Workflow

  1. Read the fast recovery target from the module plan.
  2. Read the runtime handoff and decide whether full neural recovery is available.
  3. If full recovery is blocked and soft mode allows proxy recovery, declare the reduced target.
  4. Invoke generated helper scripts for sequential proposal records, posterior transformation, and atomic loss.
  5. Run an executable experiment command that computes numeric metrics and mechanism checks.
  6. Write a source manifest that lists allowed sources and excludes original repositories.
  7. Run the Distiller recovery gate after the command completes.

Validation

The script scripts/check_recovery_artifacts.py performs a lightweight local check for key files and mechanism fields. The Distiller gate remains authoritative:

python scripts/check_recovery_artifacts.py --attempt-dir <attempt_dir>

The same script also checks that source_manifest.json does not list forbidden sources and that it mentions runtime_handoff.json, which catches common source-boundary mistakes before full analysis.

Limitations

  • This skill does not choose ambiguous paper sources.
  • It cannot turn a proxy experiment into hard-mode success.
  • It should not silently create data from a previous attempt; every recovery run must produce its own command-generated artifacts.

04 同仓库更多 Skills

VectorSpaceLab
VectorSpaceLab/AREX-Skill

passnet-triton-opt

Make a single PassNet Triton kernel fast AND numerically faithful: performance model (when a replacement can win at all), block/grid/warp tuning, autotune policy, launch overhead, and per-op numeric recipes to pass the dtype baseline tolerances. Use when a pass matches and is correct but speedup below expected, or when correctness fails by small numeric margins.

★ 351
VectorSpaceLab
VectorSpaceLab/AREX-Skill

passnet-solve

END-TO-END playbook for solving one PassNet sample: analyze the computation graphs, decide the optimization strategy (what to fuse, what to replace, what to leave alone), drive the iteration loop, and maximize the sample score. This is the ENTRY skill — invoke it first for any PassNet optimization task; it tells you when to use passnet-pattern-fusion, passnet-triton-opt and passnet-feedback.

★ 351
VectorSpaceLab
VectorSpaceLab/AREX-Skill

passnet-skill

PassNet GPU kernel optimization via compiler passes. Design and implement Triton-based optimization passes, create pass files under ./pass_dir/, self-evaluate with pass_evaluator, and iterate to maximize GPU speedup.

★ 351
VectorSpaceLab
VectorSpaceLab/AREX-Skill

passnet-pattern-fusion

Author PassNet pass files: write patterns that actually MATCH the FX graph, pick fusion regions (which ops to absorb into one Triton kernel), and structure multi-pass submissions with the mandatory shared-dispatch architecture. Use when creating or fixing pattern()/replacement_args()/replacement_func() files, when a pass "failed to match", or when deciding how to fuse multiple kernels/ops.

★ 351
VectorSpaceLab
VectorSpaceLab/AREX-Skill

passnet-orchestrate

Round/batch planning and decision support for PassNet graph optimization. Use for multi-sample triage, worker allocation, eval-budget policy, and post-eval keep/revert decisions. For solving one concrete sample end to end, use passnet-solve as the entry skill; consult this skill only when you need broader planning or a specific decision gate.

★ 351
VectorSpaceLab
VectorSpaceLab/AREX-Skill

passnet-feedback

Fast, deterministic feedback for PassNet work: pre-flight pattern/match verification WITHOUT burning a GPU evaluation, per-node bottleneck analysis of a sample's graphs, and eval-log parsing into per-variant status + estimated score + failure classification. Use BEFORE every GPU evaluation (check_pattern), at the START of a sample (analyze_graph), and AFTER every evaluation (parse_eval_log).

★ 351