`tamarillo` — coding harness inspection

In the past 2 years coding harnesses (and even the term itself) became ubiquitous. At Tamarillo one goal is to systematize the utilization of coding harnesses (that is why the theta-spec and theta were created). ~400K public GitHub repositories containing configuration files for AI coding assistants (harnesses) were fetched. ~400K was the count at time of collection after exhaustively searching GitHub public repos^[†]. The process to get the data is pretty straightforward

filter criteria: PATTERNS per harness were defined (explained in detail in appendix A)
repo search: Code searches on GitHub's REST API filtered against harness configuration files
enriching stage: GitHub's GraphQL API was used to enrich files with commit count, file bytes, creation date, etc.

This document covers a couple of things, market share and adoption dynamics, configuration surface anatomy (what files exist, how big, how often touched), multi-harness co-occurrence, repo demographics by stars/language/owner type, and other yerbas. It was created with the intent of being a DIY-thermometer for a slice of a domain. Although some suspicions were confirmed (maybe some obvious ones it MAY be argued), it is strongly suggested to the reader to read the limitations and methodology section in this document. Only public repositories were fetched. The dataset reflects configuration intentions (i.e. a repo that has a .cursorrules file signals that someone set it up, not necessarily that Cursor is daily used). This is a lower bound on harness adoption. [†] the number of repositories fetched does include a negligible amount of forks of already captured repos (<0.1%). This were excluded for the analysis.

At the time of writing, the C4H (Claude Code, Codex, Copilot, Cursor, Hermes) dominates ~80% of the public repositories that expose harness configurations. This is the Pareto principle in action — an expected outcome given that power laws are usually at play in software and software tools. From ~20 harnesses, 5 dominate.

C4H domination

Throughout the years and months, the landscape changed. At the beginning the market share was pretty much dominated by Cursor. It is no surprise to confirm that, as well as the rise of Claude Code and shortly after Codex and Hermes. Below, two charts measure the following signals: recent adoption dynamics, and the current state of market share^[†].

Rolling market share chart, showing the signal of recent adoption. That is, popularity calculated as the share for new repositories created in an x-day window, filtered by the date each harness config format became publicly available.
Evolution of market share in absolute terms over the past ~2 years.

The dates used as filters correspond to harness-specific events associated with the release date of configuration files deemed mandatory in the context of a coding harness && || the product launch date. Details and sources for each date can be found here. [†] This analysis is constrained to what GitHub's public APIs expose. Private repos, enterprise installations, and home-directory config never reach the index, so the curves here are ballpark estimates that can't be confirmed at the population level.

As stated above, the evolution of market share in absolute terms over the past ~2 years filters by release date without including the retroactive configured repos. The results displayed below are the cumulative share for each harness present in the selected group.

The cumulative adoption over time, filtered by harness release date, shows that:

All growth shows exponential behavior.
Harnesses experience rate discontinuity, e.g. Cursor ~2025-07, Copilot ~2025-11 (i.e. subtle change, but noticeable with eyesight on the rate of growth, the cause is up for elucubration)
Even though harnesses like Claude Code or Codex were released later, they already caught up in terms of repo footprint.^[†].

[†] Repo footprint may underestimate true adoption — see methodology & limitations.

Codex & Claude Code

There is colossal competition between frontier labs in many domains including harnesses, like Codex and Claude Code. The market share ratio was discussed on this post, which puts Codex at ~5% of Claude Code's usage in Sep 2025 and ~40% by Jan 2026 (emphasis on the approximation operator ~). The public-repo measurement here sits above the ~5% and pretty accurately on ~40%: two effects MAY compound to produce the gap:

WIRED reports "5 percent as much use" (Sep 2025) and "40 percent of Claude Code's user base" (Jan 2026), attributed to "people with direct knowledge of the matter" — anonymous internal sources, no methodology disclosed. This chart counts public repos with a config file committed.
Configs MAY be committed weeks or months after the user actually adopted the tool.

The proxy is more credible for cumulative adoption than for short-run growth. A 14-day rolling window is noisy and there is no public data or ballpark estimate to contrast that rate-of-change series against.

Multi-harness adoption

If each repo independently decides to adopt one more harness with constant probability

p

, the count at

k

follows a geometric distribution:

N (k) \propto p^{k}

. It MAY be a bit of a stretch, but there is no resistance to the temptation of fitting a line on a

lo g (y)

vs

x

chart and interpreting what that constant probability means.

x

is indexed over the count of co-occurrences of harness configurations in repos, so

p

is the probability of adopting harness

i + 1

after already having

i

. Note that it does not matter which harness is adopted (e.g. maybe it is always Claude Code because it is popular), what matters is that there is a new one. The decision to add harness

k + 1

does not depend on how many are already configured. The regression was done for harness counts

k \in [1, 5]

because above that the counts drop to single digits.

\overset{p}{^}

is obtained via OLS on

lo g N (k) = lo g C + k lo g p

. The 95% CI uses the textbook normal-theory slope CI

\hat{β} \pm 1.96 \cdot SE (\hat{β})

and is then pushed through

\overset{p}{^} = e^{\hat{β}}

. The regression on

lo g (N)

is heteroscedastic (small counts are noisier), so the reported CI is slightly optimistic. A Poisson GLM fitting

N (k) \sim Poisson (μ_{k}), lo g μ_{k} = lo g C + k lo g p

is the correct way of doing things. ^[†] [†] If only 2 harness are selected the linear regression will be evidently perfect.

Configuration surface anatomy

Configuration surfaces are defined as the harness-specific resources that modify the behavior of the coding agent, e.g. system prompts or mcp.json MCP tool configuration files. Every harness exposes a different set of configuration surfaces. As the ecosystem evolves and matures, a certain degree of consistency and convergence is seen. That is, many harnesses share configurations because at the end of the day they share functionality, or expose functionally equivalent behavior that is configured similarly.

File counts and memory footprint per surface category

Every matched config file is assigned to one of 12 surface categories based on what it does:

system_prompt: CLAUDE.md, .cursorrules, copilot-instructions.md, HERMES.md, SOUL.md, AGENTS.md. One per repo, usually at root.
rule: scoped rule files that apply only to specific paths or file globs: .cursor/rules/*.mdc, .github/instructions/*.instructions.md.
agent: sub-agent: .claude/agents/*.md, .codex/agents/*.toml.
skill: files following the Agent Skills standard: SKILL.md inside a skills/ directory.
prompt: reusable prompt snippets / templates: .github/prompts/*.prompt.md, .continue/prompts/, .pi/prompts/.
workflow: multi-step recipes the agent can execute: .windsurf/workflows/*.md, .clinerules/workflows/*.md.
tool_mcp: Model Context Protocol server configs: .mcp.json, .vscode/mcp.json, .cursor/mcp.json, .codex/mcp.json.
hook: lifecycle event handlers: .cursor/hooks.json, .windsurf/hooks.json, .github/hooks/*.json, .codex/hooks.json.
plugin: packaged extensions: .claude-plugin/plugin.json, .cursor-plugin/plugin.json, .opencode/plugins/.
ignore: files telling the agent what to skip: .aiderignore, .cursorignore, .codeiumignore.
settings: everything else that configures behavior: config.yaml, settings.json, MEMORY.md, USER.md, model selection, etc.
legacy: older command-style files that have largely been superseded by skills/agents: .claude/commands/*.md, .opencode/commands/*.md.

Surface decomposition per harness

For every harness, how do its config files break down across surface categories. The search pattern set is not exhaustive for every harness. Harnesses with fewer patterns will appear more concentrated in the surfaces that are covered. Percentages reflect the distribution within matched files, not the true distribution of all config files a harness supports. See methodology & limitations.

Y axis:

shareabsolute count

File size distribution by surface

A surface category can be selected below to see file-size distributions per harness. Each row is a horizontal boxplot on a log byte-size x-axis (box = IQR, white line = median, whiskers =

1.5 \times IQR

, dots = sampled files). Y-positions and colors are kept across surface switches. Some harnesses go empty for some surfaces^[†].

[†] Two things explain why a harness MAY be empty for some surface.

not every surface is available for every harness
some harnesses dump everything into one file instead of separate ones

Full pattern set in appendix A.

Per-repo configuration footprint

Total config bytes per repo — how heavy is the harness setup?

Top 25 repos by config-file count

full_name	n_files	harnesses_in	surfaces_in	stars	language	owner_type
tomevault-io/gemini-extensions	4585	rest	plugin, system_prompt	0		Organization
Brmbobo/Web2podcast	4398	claude, rest	agent, system_prompt, tool_mcp	1	Python	User
tomevault-io/cursor-plugins	3704	cursor	rule	0		Organization
Nelsonmbigili/Replication_MigrateLib	2584	copilot	prompt	0	HTML	User
alpgul/Cursorrules-Database	2056	cursor	rule, system_prompt	1	TypeScript	User
netbarros/psique	1523	codex, rest	skill, system_prompt	6	JavaScript	User
tomevault-io/copilot-plugins	1513	copilot	system_prompt	0		Organization
ZTech-Inc/ZTech_Agents	1431	hermes, rest	settings, system_prompt	0	Python	Organization
engremran07/gsmvault	1151	claude, codex, copilot, rest	agent, legacy, prompt, rule, settings, skill, system_prompt	0	Python	User
udbfd68-cell/AURION-APP	1125	claude, codex, hermes, rest	settings, skill, system_prompt	0	Python	User
blacktop/ipsw-diffs	1087	hermes, rest	settings, system_prompt	687		User
kohebth/agent-skill-sets	1058	claude, rest	skill	0	Python	User
michellab/paramfit-tests	1057	cursor	rule	0	Roff	Organization
a5c-ai/babysitter	1040	claude, codex, cursor, rest	agent, other, plugin, settings, system_prompt	585	JavaScript	Organization
fabioeducacross/DesignSystem-Vuexy	972	claude, copilot, cursor, rest	skill, system_prompt	0	Python	User
nguoikhongten02022005-cell/doan3-webquanlynhahang	957	claude, codex, copilot	agent, skill, system_prompt	0	Python	User
mdegans/agora-agents	936	hermes	system_prompt	0	Rust	User
harborgrid-justin/white-cross	874	claude, rest	agent, system_prompt	1	TypeScript	User
HIDORAKAI002/ai-workspace-archive	853	claude, codex, copilot, cursor, hermes, rest	agent, ignore, other, plugin, prompt, rule, settings, system_prompt, tool_mcp	7	TypeScript	User
ewail/FreeMat	826	cursor, hermes	rule, settings	9	HTML	User
Hariharahari/SEL	823	rest	system_prompt	0		User
Hariharahari/Agents	822	rest	system_prompt	0		User
agency-black/Blackmind	791	claude, codex, rest	skill, system_prompt, tool_mcp	0	Python	Organization
openclaw/skills	791	claude, codex, copilot, hermes, rest	other, plugin, prompt, rule, settings, system_prompt, tool_mcp	4208	Python	Organization
rudironsoni/Synaxis	754	claude, codex, copilot, cursor, rest	ignore, legacy, prompt, rule, settings, skill, system_prompt, tool_mcp	2	C#	User

harness agnostic power-law in config byte footprint $α$

It is equally impossible to resist the need of fitting a line in a log-log chart. The CCDF curves above appear roughly straight on log-log, suggesting

P (bytes \geq x) \sim x^{- α}

.

α

is fitted in the range [20 KB, 5 MB] with uncertainty estimated via bootstrap (200 resamples of the repos, not the CCDF points — because CCDF points are correlated order statistics). The results show a consistent

α

amongst harnesses. The harness agnostic

α \approx 1.3

describes a Pareto-ish tailed distribution. This speaks again to the pressenece of power laws in software and the preferential attachment dynamics on harness configuration files.

Config churn — write-once or iterate?

commit_count per file indicates how many times each config file was touched. The following can be seen

Claude Code in comparison has high iteration of its system prompt in comparison to other harnesses
Rule and Skill config churn is qualitatively consistent amongst harnesses
Most configuration files are commited and never changed (i.e. 1 commit). This MAY speak to the relaxed and toy project nature of many of the repos that were fetched.

Popularity vs stars

Star count is increasingly becoming a less trustworthy signal of quality of code. Stars SHOULD probably be interpreted as as a proxy for repo prominence, not as a sorting metric of code quality Two things are measured below

Star CCDF per harness against the all-adopters baseline.
Adoption rate across star buckets. inversing the condition: given the repo bucket (0 stars, 1–9, …, 10k+), what is the probability each harness is configured.

Star CCDF per harness

Codex shows dominance on the tails of star distribution. This MAY be attributed to maturity of userbase and this MAY be related to inherent quality of the harness. In the fake star landscape of 2026 (people buy stars in the dark market) there is one harness showing signs of dominance when measured against this proxy signal. Assuming mature use, software quality, is still positively correlated with stars, the Codex phenomenon is super interesting but beyond the scope of this analysis.

Star percentiles per harness

harness	n	P75	P90	P95	P99
all adopters	396048	1	6	26	692
claude	121744	1	4	20	505
cursor	62402	1	5	24	670
codex	61814	2	21	125	3839
copilot	58576	1	5	23	994
hermes	49204	3	22	94	1841
rest	115065	0	5	22	692

6 Star buckets categories were created to analyze the prevalence of harnesses. The categories are groups of OOM ranges. The adoption share shows a skew for Codex towards high-star repos. This was already visible in the CCDF slower death for Codex in particular. The reasons for said skew goes beyond the scope of this analysis but are definitely a trigger for future research.

Language and organizational signal

This section slices the dataset and categorizes based on langauge and on type of user.

User vs organization differential

The two type of configurators for harnesses are individual users and organizations as per the github definition. It is interesting to analyze 2 simple things

raw adoption rate for each owner type,
the log-ratio ^[†] to see the skew.

[†] The right panel shows

lo g_{2} (P_{org} / P_{user})

— the pointwise KL divergence measured in bits.

Language analysis

The interactive controls below — harness selection on the sidebar and the language dropdown — drive the two charts in this subsection.

Language rarity per harness. For the chosen language, $lo g_{2} (\frac{P ( lang ∣ harness )}{P ( lang ∣ all adopters )})$ to measure degree of indexation into a specific language.
Language mix per harness. Dodged bars showing $P (lang ∣ harness)$ across the top $K$ languages.

Repo vitality & evolution

Are these repos still maintained? How are configuration practices changing over time?

Is the repo still alive?

CCDF of days-since-push + alive-share bar.

Methodology & limitations

How the dataset was assembled

The pipeline runs in three stages.

The first stage is pattern definition: for every harness in scope, a set of non exhaustive filename and path patterns are written. The full pattern list per harness is in appendix A.
The second stage is repo search. Each pattern is fired against GitHub /search/code. GitHub caps any single query at 1000 results, queries were therefore adapted to be size-sharded by file byteSize until every shard returns under that cap. The output is one row per matched (repo, file) pair carrying full name, harness, label, path, and the blob SHA.
The third stage is a two-phase GraphQL enrichment.
- Repository query per repo to pull stars, forks, primary language, owner type (User or Organization), createdAt, pushedAt, default-branch commit total, topics and license.
- Queries each file individually for Blob.byteSize plus the most recent commit, yielding per-file size, last_commit_date, and commit_count.

Time anchoring

Two date columns define the time dependant charts. repo_created_at (the date the repository itself was created on GitHub) anchors every cumulative and rolling chart — rolling share, the two absolute-share stackplots (config-file anchor and product launch anchor), the Codex-vs-Claude ratio chart, and the surface-category evolution. Each curve is clipped at the harness's HARNESS_RELEASE (config-file date) or HARNESS_LAUNCH (product-launch date) so the cumulative count never credits a harness for repos that were created before it existed. last_commit_date (the most recent commit touching a specific config file) anchors the per-file recency and churn views — push recency, alive-share, config churn — falling back to repo_created_at when GraphQL does not return a path-scoped commit.

Limitations

Caveats that are flagged in footnotes throughout the document, consolidated here:

Public repositories only. Private repos, enterprise installations, and home-directory configuration (~/.cursor/, ~/.claude/, ~/.codex/) are out of reach. Every curve is a lower bound on real adoption.
Configuration intent is not active use. A repo with .cursorrules proves that someone set Cursor up at least once. Whether the harness is still being used daily is not observable from the file's existence.
Pattern coverage is non-exhaustive. Each harness is covered by 2–11 patterns targeting known paths. Anything stored outside the repo tree, IDE-managed state, or runtime-generated artifacts is missed. Harnesses whose configuration surface lives primarily outside the repo (e.g. settings that only exist in the home directory) are systematically under-counted.
Filename-only matching. Patterns match on path and filename, not content. filename:MEMORY.md matches a Hermes memory store and an unrelated MEMORY.md in some other repo equally; harness attribution is pattern matched, not verified through reading and analyzing content.
Cross-tool attribution is a convention. The six shared files listed above are each credited to a single canonical owner. When several harnesses in the same repo would read the same file, only the canonical owner gets the count.
Forks and templates are excluded at the source. A negligible (<0.1%) slice of the originally fetched repositories was filtered out before any aggregation.

Appendix A: search patterns per harness

Every pattern used in the GitHub /search/code sweep. Each row is one query, the harness + label pair determines how the found file is classified. Content-grep patterns (*-content) match keyword presence inside files, not filenames. Data collected: 2025-05-05. Forks and templates are excluded. Two date columns are recorded per harness. Config-file release is when the harness's specific config file format became usable in repositories — that is the date used in the rolling-share and cumulative charts. Product launch is when the harness product itself first became publicly available, which can be much earlier. For example, Cursor launched ~March 2023 but .cursorrules only appeared in April 2024. Copilot has been around since June 2021 but copilot-instructions.md only appeared in September 2024. The methodology section discusses why the config-file date is the cleaner signal for the rolling-share charts, and the product-launch chart below mirrors the catch-up chart against the launch dates.

aider (5 patterns)

aider — 5 patterns · docs product launched 2023-08-22 — earliest non-yanked PyPI release, v0.13.0 (source) config file available since 2023-06-08 (.aider.conf.yml) — PyPI v0.5.0 (source)

label	query
`config-yml`	`filename:.aider.conf.yml`
`aiderignore`	`filename:.aiderignore`
`model-settings`	`filename:.aider.model.settings.yml`
`model-metadata`	`filename:.aider.model.metadata.json`
`conventions-md`	`filename:CONVENTIONS.md`

amazonq (6 patterns)

amazonq — 6 patterns · docs product launched 2023-11-28 — Amazon Q announced at AWS re:Invent (source) config file available since 2025-03-26 (AmazonQ.md) — agent mode (source)

label	query
`system-prompt`	`filename:AmazonQ.md`
`rules`	`filename:.md path:.amazonq/rules`
`config`	`filename:settings.json path:.amazonq`
`mcp`	`filename:mcp.json path:.amazonq`
`agents`	`filename:.json path:.amazonq/cli-agents`
`prompts`	`filename:.md path:.amazonq/prompts`

augment (8 patterns)

augment — 8 patterns · docs product launched 2024-08-01 — Augment Code out of stealth ("Introducing Augment" post) (source)

label	query
`guidelines`	`filename:.augment-guidelines`
`rules`	`filename:.md path:.augment/rules`
`skills`	`filename:SKILL.md path:.augment/skills`
`commands`	`filename:.md path:.augment/commands`
`settings`	`filename:settings.json path:.augment`
`plugin`	`filename:plugin.json path:.augment-plugin`
`marketplace`	`filename:marketplace.json path:.augment-plugin`
`augmentignore`	`filename:.augmentignore`

claude (10 patterns)

claude — 10 patterns · docs product launched 2025-02-24 — Claude Code research preview (source) config file available since 2025-02-24 (CLAUDE.md) — research preview (source)

label	query
`system-prompt`	`filename:CLAUDE.md`
`settings`	`filename:settings.json path:.claude`
`settings-local`	`filename:settings.local.json path:.claude`
`rules`	`filename:.md path:.claude/rules`
`agents`	`filename:.md path:.claude/agents`
`commands`	`filename:.md path:.claude/commands`
`skills`	`filename:SKILL.md path:.claude/skills`
`output-styles`	`filename:.md path:.claude/output-styles`
`loop`	`filename:loop.md path:.claude`
`plugin`	`filename:plugin.json path:.claude-plugin`

cline (10 patterns)

cline — 10 patterns · docs product launched 2024-07-01 — original 'claude-dev' VS Code extension, mid-2024 (month-precision) (source) config file available since 2025-01-14 (.clinerules/) — v2.x (source)

label	query
`clinerules`	`filename:.clinerules`
`clineignore`	`filename:.clineignore`
`roomodes`	`filename:.roomodes`
`roorules`	`filename:.md path:.roo/rules`
`roo-mcp`	`filename:mcp.json path:.roo`
`rooignore`	`filename:.rooignore`
`skills`	`filename:SKILL.md path:.cline/skills`
`rules-dir`	`filename:.md path:.clinerules`
`roo-skills`	`filename:SKILL.md path:.roo/skills`
`clinerules-workflows`	`path:.clinerules/workflows`

codex (7 patterns)

codex — 7 patterns · docs product launched 2025-04-16 — Codex CLI open-sourced (source) config file available since 2025-04-16 (AGENTS.md) — open-sourced (source)

label	query
`config`	`filename:config.toml path:.codex`
`agents-override`	`filename:AGENTS.override.md`
`agents-toml`	`filename:.toml path:.codex/agents`
`rules-starlark`	`filename:.rules path:.codex/rules`
`hooks`	`filename:hooks.json path:.codex`
`mcp`	`filename:mcp.json path:.codex`
`skills`	`filename:SKILL.md path:.codex/skills`

continue (10 patterns)

continue — 10 patterns · docs product launched 2023-05-27 — VS Code marketplace first release (source) config file available since 2023-08-10 (config.json) — VS Code extension first public (source)

label	query
`config-yaml`	`filename:config.yaml path:.continue`
`config-json`	`filename:config.json path:.continue`
`rules`	`filename:.continuerules`
`rules-dir`	`filename:.md path:.continue/rules`
`prompts`	`filename:.md path:.continue/prompts`
`prompts-legacy`	`filename:.prompt path:.continue/prompts`
`continueignore`	`filename:.continueignore`
`skills`	`filename:SKILL.md path:.continue/skills`
`mcp-servers`	`filename:.json path:.continue/mcpServers`
`settings`	`filename:settings.json path:.continue`

copilot (7 patterns)

copilot — 7 patterns · docs product launched 2024-09-19 — copilot-instructions.md introduced (harness-config-relevant date; product launched 2021-06-29) (source) config file available since 2024-09-19 (copilot-instructions.md) — VS Code 1.94 (source)

label	query
`system-prompt`	`filename:copilot-instructions.md`
`instructions`	`filename:instructions.md path:.github/instructions`
`agents`	`filename:.agent.md path:.github`
`skills`	`filename:SKILL.md path:.github/skills`
`prompts`	`filename:.prompt.md`
`hooks`	`filename:.json path:.github/hooks`
`setup-steps`	`filename:copilot-setup-steps.yml path:.github/workflows`

cursor (10 patterns)

cursor — 10 patterns · docs product launched 2023-03-01 — Cursor IDE first released by Anysphere (Wikipedia initial release: 2023; month-precision) (source) config file available since 2024-04-12 (.cursorrules) — v0.32.x changelog (source)

label	query
`rules-md`	`filename:.md path:.cursor/rules`
`rules-mdc`	`filename:.mdc`
`cursorrules`	`filename:.cursorrules`
`agents`	`filename:.md path:.cursor/agents`
`mcp`	`filename:mcp.json path:.cursor`
`hooks`	`filename:hooks.json path:.cursor`
`skills`	`filename:SKILL.md path:.cursor/skills`
`cursorignore`	`filename:.cursorignore`
`plugin`	`filename:plugin.json path:.cursor-plugin`
`marketplace-manifest`	`filename:marketplace.json path:.cursor-plugin`

gemini (8 patterns)

gemini — 8 patterns · docs product launched 2025-06-23 — gemini-cli early-access tag (source) config file available since 2025-06-25 (GEMINI.md) — CLI launch (source)

label	query
`system-prompt`	`filename:GEMINI.md`
`settings`	`filename:settings.json path:.gemini`
`geminiignore`	`filename:.geminiignore`
`commands`	`filename:.toml path:.gemini/commands`
`skills`	`filename:SKILL.md path:.gemini/skills`
`agents`	`filename:.md path:.gemini/agents`
`policies`	`filename:.toml path:.gemini/policies`
`extension`	`filename:gemini-extension.json`

goose (6 patterns)

goose — 6 patterns · docs

label	query
`instructions`	`filename:.goosehints`
`skills`	`filename:SKILL.md path:.goose/skills`
`agents`	`filename:.md path:.goose/agents`
`recipes`	`filename:.yaml path:.goose/recipes`
`recipes-json`	`filename:.json path:.goose/recipes`
`gooseignore`	`filename:.gooseignore`

hermes (10 patterns)

hermes — 10 patterns · docs product launched 2026-02-25 — Hermes Agent release per Nous Research releases page (source) config file available since 2025-02-20 (SOUL.md) — initial release (source)

SOUL.md originated in the openclaw/hermes lineage (soul.md, openclaw docs) and is not documented as a config file by any other harness. MEMORY.md and USER.md are hermes conventions but the filenames are generic. Without content analysis, matches cannot be distinguished from unrelated documentation files that share the same name.

label	query
`system-prompt`	`filename:HERMES.md`
`system-prompt-dot`	`filename:.hermes.md`
`soul`	`filename:SOUL.md`
`memory`	`filename:MEMORY.md`
`user-profile`	`filename:USER.md`
`cli-config`	`filename:cli-config.yaml`
`cli-config-example`	`filename:cli-config.yaml.example`
`plans`	`path:.hermes/plans extension:md`
`skills`	`filename:SKILL.md path:.hermes/skills`
`personalities`	`path:.hermes/personalities extension:md`

jetbrains (8 patterns)

jetbrains — 8 patterns · docs product launched 2023-12-06 — JetBrains AI Assistant public preview (source)

label	query
`junie-agents-md`	`filename:AGENTS.md path:.junie`
`ai-rules`	`filename:.md path:.aiassistant/rules`
`aiignore`	`filename:.aiignore`
`junie-guidelines`	`filename:guidelines.md path:.junie`
`noai`	`filename:.noai`
`review-guidelines`	`filename:review-guidelines.md path:.ai`
`review-rules`	`filename:review-rules.md path:.ai`
`review-guidelines-root`	`filename:REVIEW_GUIDELINES.md`

opencode (8 patterns)

opencode — 8 patterns · docs product launched 2025-05-14 — earliest tagged release (v0.0.45) (source) config file available since 2025-01-15 (opencode.json) — initial release (source)

label	query
`config-json`	`filename:opencode.json`
`config-jsonc`	`filename:opencode.jsonc`
`agents`	`filename:.md path:.opencode/agents`
`skills`	`filename:SKILL.md path:.opencode/skills`
`commands`	`filename:.md path:.opencode/commands`
`tools`	`path:.opencode/tools`
`plugins`	`path:.opencode/plugins`
`package-json`	`filename:package.json path:.opencode`

pi (6 patterns)

pi — 6 patterns · docs product launched 2025-10-01 — pi-mono / earendil-works/pi agent harness, ~Q4 2025 per lucumr blog (month-precision) (source) config file available since 2025-03-15 (PI.md) — initial release (source)

label	query
`system-prompt`	`filename:SYSTEM.md path:.pi`
`append-system-prompt`	`filename:APPEND_SYSTEM.md path:.pi`
`settings`	`filename:settings.json path:.pi`
`skills`	`filename:SKILL.md path:.pi/skills`
`extensions`	`path:.pi/extensions`
`prompts`	`filename:.md path:.pi/prompts`

supermaven (2 patterns)

supermaven — 2 patterns · docs product launched 2024-04-01 — Supermaven 1.0 (month-precision) (source)

shallow configuration surface — limited pattern coverage.

label	query
`config`	`filename:supermaven.json`
`rules`	`filename:.supermaven-rules`

tabnine (2 patterns)

tabnine — 2 patterns

shallow configuration surface — limited pattern coverage.

label	query
`config`	`filename:.tabnine.json`
`rules`	`filename:tabnine-rules.json`

trae (4 patterns)

trae — 4 patterns · docs product launched 2025-01-20 — v1.0.0 beta (source) config file available since 2025-01-20 (.trae/rules/) — v1.0.0 beta (source)

label	query
`rules`	`filename:.md path:.trae/rules`
`mcp`	`filename:mcp.json path:.trae`
`skills`	`filename:SKILL.md path:.trae/skills`
`ignore`	`filename:.ignore path:.trae`

void (2 patterns)

void — 2 patterns · docs product launched 2025-01-18 — voideditor v1.0.0 beta (source)

shallow configuration surface — limited pattern coverage.

label	query
`config`	`filename:settings.json path:.void`
`rules`	`filename:.void-rules`

windsurf (7 patterns)

windsurf — 7 patterns · docs product launched 2024-11-12 — Wave 1 launch (source) config file available since 2024-11-12 (.windsurfrules) — Wave 1 launch (source)

label	query
`rules`	`filename:.md path:.windsurf/rules`
`windsurfrules`	`filename:.windsurfrules`
`skills`	`filename:SKILL.md path:.windsurf/skills`
`workflows`	`filename:.md path:.windsurf/workflows`
`hooks`	`filename:hooks.json path:.windsurf`
`codeiumignore`	`filename:.codeiumignore`
`deployment`	`filename:windsurf_deployment.yaml`

zed (3 patterns)

zed — 3 patterns · docs product launched 2024-08-20 — Zed AI assistant panel launch (source)

shallow configuration surface — limited pattern coverage.

label	query
`settings`	`filename:settings.json path:.zed`
`tasks`	`filename:tasks.json path:.zed`
`rules`	`filename:.rules`

tamarillo — coding harness inspection

Harness adoption: market share

C4H domination

Rolling share — new harness adoptions

Codex & Claude Code

Multi-harness adoption

Configuration surface anatomy

File counts and memory footprint per surface category

Surface decomposition per harness

File size distribution by surface

Per-repo configuration footprint

harness agnostic power-law in config byte footprint α

Config churn — write-once or iterate?

Popularity vs stars

Star CCDF per harness

Adoption share across star buckets

Language and organizational signal

User vs organization differential

Language analysis

Repo vitality & evolution

Is the repo still alive?

Methodology & limitations

How the dataset was assembled

Time anchoring

Limitations

Appendix A: search patterns per harness

`tamarillo` — coding harness inspection

harness agnostic power-law in config byte footprint $α$