Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Check for updates and upgrade Ouroboros to the latest version
.claude/skills/q00-update/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-06 | ✗→✓ | ▲ Improved | 39% | 0% |
| case-17 | ✗→✓ | ▲ Improved | -2% | 0% |
| case-04 | ✗→✓ | ▲ Improved | 56% | 0% |
| case-11 | ✗→✓ | ▲ Improved | -6% | 0% |
| case-13 | ✗→✓ | ▲ Improved | 19% | 0% |
Check for updates and upgrade Ouroboros without changing the installation's manager, environment, or optional-dependency profile.
ooo update
/ouroboros:updateTrigger keywords: "ooo update", "update ouroboros", "upgrade ouroboros"
When the user invokes this skill:
bash ouroboros update --check
now or Skip.
bash ouroboros update --yes
When the user actively uses both Claude Code and Codex, refresh both host integrations without changing the configured execution backend:
bash ouroboros update --yes --runtime all
Show the command's result. The native updater binds the package upgrade and post-update setup to one receipt-backed installation identity: manager, environment, recorded profile, and environment-local console script.
ouroboros update --help works,report the native error and stop. Do not replace it with a manual updater. In particular, do not infer ownership from global uv tool list, pipx list, PATH order, directory names, or the active agent runtime.
ouroboros update --help does not work, this is a legacy installationthat predates the receipt-bound updater. Fail closed:
ouroboros --version check is allowed.environment, and extras/profile, so it will not mutate the installation.
ouroboros-ai[...] extras and custom manager root) to reach a version with the native updater.
install rather than guessing.
legacy path.
the native command. If project instruction content also needs regeneration, suggest ooo setup; do not edit project instruction files as part of the package update.
uv and pipx upgrades replay the running environment's local receipt,preserving base, [tui], [mcp,tui], [claude,tui], [all], and other recorded profiles.
OpenCode plugin/subprocess topology instead of inferring a replacement from PATH.
then the persisted orchestrator.*_cli_path, then PATH. The chosen exact executable is validated and reused for plugin/setup refresh so a stale PATH binary cannot replace an operator-selected runtime.
--runtime all refreshes the Claude and Codex plugin integrations plusinstalled runtime artifacts without changing the configured execution backend. Active Codex sessions still need a restart because Codex does not currently retain an in-use plugin generation; Claude may use /reload-plugins or restart.
ouroboros update supports --check, --yes, --dry-run,--prerelease, and --runtime; see ouroboros update --help.
Your final response MUST end with exactly one breadcrumb footer line:
◆ <current state> → next: <recommended action>Derive <current state> from live session state via ouroboros_session_status when that MCP projection is available; otherwise derive it from this skill's actual outcome. Never use a linear Step N of M footer because Ouroboros is an evolutionary loop. When the next action is genuinely a choice, list 2-3 honest options in the next: clause. The breadcrumb line must be the last line of the response.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-06 | fail→pass | 9,708 | 6,814 | -30% | 1 | 1 | 0% | 1,613 | 2,247 | +39% | 0 | 0 | — |
case-07 | fail→fail | 11,785 | 11,535 | -2% | 1 | 1 | 0% | 1,687 | 2,169 | +29% | 0 | 0 | — |
case-01 | fail→fail | 9,945 | 6,468 | -35% | 1 | 1 | 0% | 1,367 | 1,233 | -10% | 0 | 0 | — |
case-17 | fail→pass | 12,682 | 3,355 | -74% | 1 | 1 | 0% | 1,564 | 1,530 | -2% | 0 | 0 | — |
case-02 | fail→fail | 9,579 | 3,934 | -59% | 1 | 1 | 0% | 1,763 | 1,204 | -32% | 0 | 0 | — |
case-03 | fail→fail | 10,866 | 5,216 | -52% | 1 | 1 | 0% | 1,873 | 1,294 | -31% | 0 | 0 | — |
case-04 | fail→pass | 12,159 | 12,284 | +1% | 1 | 1 | 0% | 1,956 | 3,053 | +56% | 0 | 0 | — |
case-05 | fail→fail | 13,140 | 5,567 | -58% | 1 | 1 | 0% | 1,886 | 1,977 | +5% | 0 | 0 | — |
case-08 | pass→pass | 12,845 | 11,121 | -13% | 1 | 1 | 0% | 2,119 | 2,058 | -3% | 0 | 0 | — |
case-09 | pass→pass | 14,309 | 4,518 | -68% | 1 | 1 | 0% | 2,043 | 1,812 | -11% | 0 | 0 | — |
case-10 | pass→pass | 12,971 | 9,366 | -28% | 1 | 1 | 0% | 1,917 | 2,697 | +41% | 0 | 0 | — |
case-11 | fail→pass | 13,905 | 4,609 | -67% | 1 | 1 | 0% | 1,979 | 1,865 | -6% | 0 | 0 | — |
case-12 | pass→pass | 11,668 | 6,546 | -44% | 1 | 1 | 0% | 1,750 | 1,889 | +8% | 0 | 0 | — |
case-13 | fail→pass | 11,429 | 7,080 | -38% | 1 | 1 | 0% | 1,890 | 2,258 | +19% | 0 | 0 | — |
case-14 | fail→pass | 10,855 | 10,954 | +1% | 1 | 1 | 0% | 2,010 | 3,043 | +51% | 0 | 0 | — |
case-15 | pass→pass | 12,572 | 5,728 | -54% | 1 | 1 | 0% | 1,880 | 2,075 | +10% | 0 | 0 | — |
case-16 | fail→pass | 16,856 | 5,950 | -65% | 1 | 1 | 0% | 2,406 | 2,172 | -10% | 0 | 0 | — |
case-18 | pass→pass | 15,649 | 6,291 | -60% | 1 | 1 | 0% | 2,054 | 2,158 | +5% | 0 | 0 | — |
case-19 | pass→pass | 14,487 | 5,997 | -59% | 1 | 1 | 0% | 1,822 | 1,904 | +5% | 0 | 0 | — |
case-20 | fail→pass | 6,629 | 9,029 | +36% | 1 | 1 | 0% | 1,112 | 2,401 | +116% | 0 | 0 | — |
case-21 | pass→pass | 13,177 | 7,905 | -40% | 1 | 1 | 0% | 1,998 | 2,564 | +28% | 0 | 0 | — |
case-22 | fail→pass | 9,100 | 8,425 | -7% | 1 | 1 | 0% | 1,238 | 2,634 | +113% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted, and 19 counted toward the lift figure. The other 3 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +41 percentage points is the difference between those two pass rates over the 19 comparable cases.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.