SkillOptimizer
Research hub
Null resultAI usage (tokens)

Helpers built from documented steps

Does moving a skill's documented creation steps into a helper reach a test at all?

measured-local2026-09-195658cf4
  • docs/experiments/COMPILER-2026-09-19.md · 5658cf4 · 2026-09-19

Headline figures

15

Skills attempted

measured-local · 2026-09-19

0of 15

Reached a three-arm comparison

measured-local · n=15 · 2026-09-19

0of 0

Accepted

measured-local · n=0 · 2026-09-19

6

Every task already solved unaided

measured-local · 2026-09-19

4

No baseline advantage

measured-local · 2026-09-19

3

Eligible, but none of the documented steps appeared in a real run

measured-local · 2026-09-19

1

Too few repetitions survived

measured-local · 2026-09-19

1

Provider error

measured-local · 2026-09-19

Notes

  • All three eligible skills were refused because none of their documented creation steps appeared in a real run: 0 of 4, 0 of 4 and 0 of 5.
  • The binding constraint is upstream of the compiler: no corpus skill of this static shape performed the steps it documents.

Tables

Helper extraction candidates, lane K

n: Runs per arm behind the row: repetitions times tasks compared, or the graded runs of the eligibility check when the skill was not compared. 15 rows.

skillnameshapeinstallsstatusstatus_supersededdecisionacceptedartifactsskill_md_unchangedtoken_reductionspread_minspread_maxstraddles_zeronet_call_changetokens_nonetokens_originaltokens_optimizedscore_nonescore_originalscore_optimizedmission_preservedretention_bandRtasksnevidence_classdatecommit
block-no-verify-hookBlock No-Verify Hookreference-lookup5,474no_baseline_advantage——No———————————————313measured-local2026-09-195658cf4666bb9158f16708aa0a48dfb3c59d752e
check-context———error——No————————————————00measured-local2026-09-195658cf4666bb9158f16708aa0a48dfb3c59d752e
code-auditCode Audit Patternsreference-lookup—no_artifact——No———————————————313measured-local2026-09-195658cf4666bb9158f16708aa0a48dfb3c59d752e
deserialization-insecureInsecure Deserializationreference-lookup3,091tasks_not_discriminatinginsufficient_repetitions—No———————————————324measured-local2026-09-195658cf4666bb9158f16708aa0a48dfb3c59d752e
fuzzing-dictionaryFuzzing Dictionaryreference-lookup4,651tasks_not_discriminating——No———————————————326measured-local2026-09-195658cf4666bb9158f16708aa0a48dfb3c59d752e
invoice-organizerInvoice Organizerreference-prose—no_baseline_advantage——No———————————————326measured-local2026-09-195658cf4666bb9158f16708aa0a48dfb3c59d752e
ios-ettrace-performanceiOS ETTrace Performancereference-lookup850insufficient_repetitions——No———————————————322measured-local2026-09-195658cf4666bb9158f16708aa0a48dfb3c59d752e
jndi-injectionJNDI Injectionreference-lookup3,059tasks_not_discriminating——No———————————————326measured-local2026-09-195658cf4666bb9158f16708aa0a48dfb3c59d752e
kernel-exploitationKernel Exploitationreference-lookup3,095no_baseline_advantage——No———————————————326measured-local2026-09-195658cf4666bb9158f16708aa0a48dfb3c59d752e
memory-forensicsMemory Forensicsworkflow9,993tasks_not_discriminating——No———————————————326measured-local2026-09-195658cf4666bb9158f16708aa0a48dfb3c59d752e
minecraft-modpack-serverMinecraft Modpack Server Setupreference-lookup—no_artifact——No———————————————313measured-local2026-09-195658cf4666bb9158f16708aa0a48dfb3c59d752e
offensive-toctouTOCTOU - Time-of-Check / Time-of-Use Exploitationreference-lookup—no_artifact——No———————————————324measured-local2026-09-195658cf4666bb9158f16708aa0a48dfb3c59d752e
ppt-editing-skillEditing Presentationsreference-lookup—no_baseline_advantage——No———————————————326measured-local2026-09-195658cf4666bb9158f16708aa0a48dfb3c59d752e
steganography-techniquesSteganography Techniquesworkflow2,994tasks_not_discriminating——No———————————————326measured-local2026-09-195658cf4666bb9158f16708aa0a48dfb3c59d752e
traffic-analysis-pcapPCAP Traffic Analysisworkflow3,145tasks_not_discriminatinginsufficient_repetitions—No———————————————324measured-local2026-09-195658cf4666bb9158f16708aa0a48dfb3c59d752e