Lifelong LLM agents increasingly rely on external skill artifacts as one element for preserving and reusing capabilities over time. These skills (usually portable Markdown files such as SKILL.md) describe when and how to apply a capability and must be corrected, expanded, and consolidated as tools and usage patterns shift over deployment. Recent work seeks to automate skill curation, but it largely evaluates against automated baselines and treats human maintenance as an unmeasured bottleneck. We study that missing process directly. We mine the full commit histories of five public AI-skill repositories, a purposive sample of AI-tooling organizations, covering 873 commits, 143 skill files, and 254 substantive post-creation edits from October 2025 to June 2026. We code each edit with pre-registered governance, operation, and trigger-evidence codebooks. Three findings emerge. First, every substantive edit is authored or merged through a named human account, while 62% carry an AI co-author trailer, with large repository-level variation. Second, these edits are genuine curation: an audited sample shows that most change skill content, and the coded operations are dominated by additions and corrections. Third, a pre-registered rule-likeness axis fails its reliability gate; reliably coding rule-likeness from commit artifacts remains an open measurement problem. We release the corpus, codebooks, mining scripts, and a replay protocol for automated skill curators. For self-evolving agents, public skill maintenance currently looks less like an autonomous pipeline than a human-governed, AI-assisted loop that future curators must measure against and operate within.
Figures & tables
Long.
Object
Curation
SkillOpt / SkillOS / Trace2Skill ( Yang et al., 2026 ; Ouyang et al., 2026 ; Ni et al., 2026 )
✓
skill files
automated only
SkillLearnBench / SWE-Skills-Bench ( Zhong et al., 2026 ; Han et al., 2026 )
–
skill files
endpoint only
Agent READMEs ( Chatlatanagulchai et al., 2025 )
✓
context files
coarse, uncoded
Claude-skills marketplace ( Ling et al., 2026 )
–
skill files
–
SkillHone ( Li & Hu, 2026 )
✓
skill files
synthetic
SkillGraph / ExpGraph ( Li et al., 2026b ; Feng et al., 2026 )
✓
internal graph
–
Table 1: Where this study sits. To our knowledge, prior work has not mined longitudinal maintenance histories of human-curated Markdown skill files with coded operations and governance. Long. = longitudinal; Object = the artifact studied; Curation = coded human-curation analysis.
repository
SKILL.md
subst. edits
edit-days
span
getsentry/skills
28
81
40
2026-01–2026-06
trailofbits/skills
74
78
22
2026-01–2026-06
obra/superpowers
14
62
26
2025-10–2026-05
anthropics/skills
18
19
11
2025-11–2026-06
cloudflare/skills
9
14
10
2026-02–2026-04
total
143
254
Table 2: The purposive corpus: per-repository skill-file count, substantive edits, distinct edit-days, and span (months). Pinned commits are listed in the reproducibility note.
Figure 1: Operations over the 254 substantive skill-file edits. Middle row: the eight coded operations (counts in the legend). Top row: the same edits grouped into the corrective and enhancement maintenance classes. Bottom row: the recount on the 165 edits that remain after excluding multi-file mass-refactor commits.
Figure 2: Human-governed maintenance; trailer-visible AI co-authorship varies by organization. Repository shares by governance class from commit metadata (counts in Table 3 ); every substantive edit is human-authored or human-merged. Classes indicate trailer-visible co-authorship, not measured AI use; squash workflows can drop trailers (§ 6 ).These repository-level shares should not be interpreted as measures of organization-wide AI adoption, human contribution, or engineering practice.
repository
n
AI
HC
HM
getsentry/skills
81
75
6
–
trailofbits/skills
78
72
1
5
obra/superpowers
62
10
52
–
anthropics/skills
19
1
–
18
cloudflare/skills
14
–
1
13
total
254
158
60
36
Table 3: Governance by repository (deterministic from commit metadata). AI = AI co-author trailer present; HC = human-committed; HM = human-merged via pull request, with no trailer-visible AI co-authorship.
Figure 3: Maintenance moves individual skills both ways, but not on average. Maintained-minus-earliest judged-quality difference (1–5 rubric). Points show skill means over 11 tasks × 3 replicates; whiskers show descriptive task-level 95% CIs; † marks deep-history skills; square marks the pre-registered skill-level aggregate. Per-skill differences are specific to this constructed transfer-task harness and should not be interpreted as measures of intrinsic skill quality, maintainer quality, or effectiveness in the repositories’ native workflows.
Appendix figures & tables4 assets
Supplementary material from the paper’s appendix.
Appendix
operation
definition
illustrative edit
distinguished from
content-expansion
adds new guidance, sections, or examples (substance)
adds a “common pitfalls” subsection
description-tuning (wording only)
factual-correction
fixes an incorrect statement, value, or API
corrects a flag name or version string
fix-from-failure (no failure cue)
restructure
reorganizes existing content
splits one section into three
content-expansion (adds substance)
fix-from-failure
responds to an observed usage failure
adds a stop condition after a reported loop
factual-correction (no failure cue)
description-tuning
rewords the name / description / trigger text
tightens the “use when” line
content-expansion (adds substance)
consolidation-merge
merges or dedupes content
folds two overlapping steps into one
restructure (no dedupe)
Appendix
Table 4: The eight curation operations with an illustrative edit and the boundary distinction that most often decides the label. The released datasheet carries full per-label examples and precedence rules.
coded operation
repository / skill
commit subject (verbatim)
lines
content-expansion
getsentry / blog-writing-guide
add bad/good examples to AI patterns section
34
factual-correction
trailofbits / solana-vulnerability-scanner
correct duplicate section numbering
15
fix-from-failure
getsentry / create-branch
Simplify Step 5 and fix unconditional stash pop
61
restructure
getsentry / agents-md
Improve agent guidance skills
147
Appendix
Table 5: Four real coded edits from the corpus, grounding the codebook in concrete cases. Subjects are quoted from the commits (conventional-commit prefixes and PR numbers elided); lines = changed lines in the named skill file.
Figure 4: Maintenance dynamics (§ 5 ). (a) Days between successive edits to a skill, per repository (488 intervals; box = IQR, whiskers 5–95%, dashed line = overall median 5 days): cadence is frequent but organization-dependent, from a 1.5-day median (obra) to 14.5 (trailofbits). (b) Net change in a skill’s resident token length between its first and last observation (120 skills with ≥2 size observations): trajectories are growth-leaning, 32 growing more than 10%, 7 shrinking more than 10%, 81 stable, so the net-change distribution is predominantly stable-to-additive, with neither universal growth nor broad decay. Recomputed from the released records; the tail is clipped at ±250% .
Table 6: Robustness battery for the powered maintenance-benefit study. The primary skill-level test (top row) is the reportable result; the next three rows are sensitivity checks; the last row reports judging-harness reliability rather than a maintenance-benefit test. No rubric dimension survives Holm correction.