Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Help the user recover lost AxonFlow tenant credentials via the email magic-link flow, or apply a paid Pro-tier license token. Use when the user says they've lost their AXONFLOW_AUTH, lost their secret, can't authenticate, or has a new AXON- license token to install.
.claude/skills/hashgraph-online-recover-credentials/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-05 | ✗→✓ | ▲ Improved | -16% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -34% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 14% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 53% | 0% |
| case-12 | ✗→✓ | ▲ Improved | -6% | 0% |
The AxonFlow agent exposes two recovery flows users sometimes need to reach from inside Codex:
AXONFLOW_AUTH / tenant secret and needs new credentials. They request a magic link, click it, copy the hex token, and the plugin persists the new credentials atomically into ~/.codex/axonflow.toml.AXON--prefixed license token by email. They paste it through the same recovery surface and the plugin persists it for the next Codex session to pick up.Use the exec_command tool to invoke the recovery script that ships with the plugin:
| User intent | Command | |---|---| | Lost credentials, request new magic link | bash $PLUGIN_DIR/scripts/recover.sh request | | Click magic link, paste hex token | bash $PLUGIN_DIR/scripts/recover.sh verify | | Install paid Pro license token | bash $PLUGIN_DIR/scripts/recover.sh apply-token | | Check current tier / config | bash $PLUGIN_DIR/scripts/recover.sh status |
$PLUGIN_DIR resolves to the plugin checkout the user installed.
For non-interactive use (test harnesses, CI, automation), the script also reads:
AXONFLOW_RECOVER_EMAIL for requestAXONFLOW_RECOVER_TOKEN for verifyAXONFLOW_LICENSE_TOKEN for apply-tokenIn each case, after the script runs, advise the user that the new credentials live in ~/.codex/axonflow.toml (mode 0600) and that they should restart Codex (or re-export AXONFLOW_AUTH / AXONFLOW_LICENSE_TOKEN from the file) for the change to take effect.
Never paste the token or secret into chat. The script reads from stdin or env so the value never enters the model context.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-05 | fail→pass | 15,551 | 5,742 | -63% | 1 | 1 | 0% | 1,736 | 1,453 | -16% | 0 | 0 | — |
case-01 | fail→fail | 12,150 | 14,508 | +19% | 1 | 1 | 0% | 1,098 | 1,087 | -1% | 0 | 0 | — |
case-02 | fail→fail | 16,514 | 6,857 | -58% | 1 | 1 | 0% | 1,671 | 889 | -47% | 0 | 0 | — |
case-03 | fail→fail | 12,171 | 13,327 | +9% | 1 | 1 | 0% | 883 | 938 | +6% | 0 | 0 | — |
case-04 | fail→fail | 19,578 | 16,700 | -15% | 1 | 1 | 0% | 1,721 | 1,348 | -22% | 0 | 0 | — |
case-06 | fail→pass | 21,973 | 5,084 | -77% | 1 | 1 | 0% | 1,453 | 964 | -34% | 0 | 0 | — |
case-07 | fail→pass | 15,305 | 21,303 | +39% | 1 | 1 | 0% | 1,479 | 1,682 | +14% | 0 | 0 | — |
case-08 | pass→pass | 17,926 | 10,598 | -41% | 1 | 1 | 0% | 1,890 | 1,245 | -34% | 0 | 0 | — |
case-09 | fail→pass | 11,434 | 14,722 | +29% | 1 | 1 | 0% | 840 | 1,285 | +53% | 0 | 0 | — |
case-10 | fail→fail | 13,507 | 54,857 | +306% | 1 | 1 | 0% | 1,137 | 3,090 | +172% | 0 | 0 | — |
case-11 | fail→fail | 10,033 | 20,048 | +100% | 1 | 1 | 0% | 959 | 1,585 | +65% | 0 | 0 | — |
case-12 | fail→pass | 10,361 | 6,183 | -40% | 1 | 1 | 0% | 1,439 | 1,355 | -6% | 0 | 0 | — |
case-13 | fail→pass | 13,448 | 5,119 | -62% | 1 | 1 | 0% | 1,293 | 1,380 | +7% | 0 | 0 | — |
case-14 | fail→pass | 13,919 | 9,761 | -30% | 1 | 1 | 0% | 1,385 | 1,291 | -7% | 0 | 0 | — |
case-15 | fail→pass | 11,726 | 3,945 | -66% | 1 | 1 | 0% | 1,951 | 951 | -51% | 0 | 0 | — |
case-16 | fail→pass | 11,892 | 7,890 | -34% | 1 | 1 | 0% | 1,100 | 848 | -23% | 0 | 0 | — |
case-17 | fail→pass | 15,456 | 4,004 | -74% | 1 | 1 | 0% | 1,654 | 1,111 | -33% | 0 | 0 | — |
case-18 | fail→pass | 10,540 | 5,908 | -44% | 1 | 1 | 0% | 1,733 | 1,582 | -9% | 0 | 0 | — |
case-19 | pass→fail | 14,792 | 13,674 | -8% | 1 | 1 | 0% | 1,423 | 1,061 | -25% | 0 | 0 | — |
case-20 | fail→pass | 19,239 | 4,581 | -76% | 1 | 1 | 0% | 2,061 | 1,245 | -40% | 0 | 0 | — |
case-21 | pass→pass | 12,985 | 11,620 | -11% | 1 | 1 | 0% | 1,642 | 1,469 | -11% | 0 | 0 | — |
case-22 | fail→pass | 18,176 | 9,124 | -50% | 1 | 1 | 0% | 1,819 | 1,196 | -34% | 0 | 0 | — |
case-23 | fail→pass | 12,316 | 8,605 | -30% | 1 | 1 | 0% | 1,891 | 1,102 | -42% | 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. 23 cases were attempted, and 16 counted toward the lift figure. The other 7 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 +57 percentage points is the difference between those two pass rates over the 16 comparable cases. 3 cases got worse with the skill loaded, and they are included in that figure.
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.