rudder-real-runtime-verifier-maintainer

Name: rudder-real-runtime-verifier-maintainer
Author: Undertone0809

247stars30forks

How to install rudder-real-runtime-verifier-maintainer?

undertone0809/rudder/rudder-real-runtime-verifier-maintainer

$npx -y skills add Undertone0809/rudder --skill rudder-real-runtime-verifier-maintainer

Installs into the current project.

Summary

Use when verifying Rudder agent runtime behavior in a real local environment, especially MCP/native Rudder tools across Codex, Claude, OpenCode, Pi, or user-named runtimes. Trigger for requests like 真是/真实环境跑过吗, 排查所有 rudder tools, transcript/fallback verification, runtime MCP avai

Files · 1

View on GitHub

SKILL.md

1	# Rudder Real Runtime Verifier Maintainer
2
3	Verify Rudder agent runtime behavior on the user's real local Rudder instance.
4	This is a black-box runtime acceptance workflow, not an implementation or code
5	review workflow.
6
7	Default to Chinese when the user asks in Chinese. Put the current truth first:
8	which runtimes passed, failed, or were blocked, and what transcript evidence
9	proves it.
10
11	## Role Boundary
12
13	Default to verification and diagnosis only:
14
15	- Run real local runtime probes and inspect run transcripts, logs, API state,
16	issue state, comments, and run metadata.
17	- Create disposable orgs, agents, issues, and runs when needed for proof.
18	- Separate provider/model failure from Rudder adapter failure.
19	- Report exact blockers and smallest likely fixes.
20	- Do not edit source files, configs, git state, or product docs unless the user
21	explicitly asks for a mutation after the verification request.
22
23	If the user asks to fix the issue, hand back to the lifecycle implementation
24	route or make the smallest explicit patch, then require this skill's real
25	runtime proof again before claiming done.
26
27	## When This Skill Wins
28
29	Use this skill when the core question is whether an agent runtime actually did
30	the work through Rudder-managed tools in a real local run.
31
32	Typical prompts:
33
34	- "你所有的 agent runtime 都本地测过跑过真实环境了吗?"
35	- "OpenCode and Pi agent 你也测了吗?"
36	- "看 transcript，别让它 fallback 用 rudder cli"
37	- "排查所有 rudder tools，都试一遍"
38	- "MCP tool 报 org id/auth 问题，正常 agent 调 tool 不该传 org"
39	- "Codex/Claude/OpenCode/Pi 真实环境跑一下"
40
41	If the user asks for general product acceptance that is not runtime/tool-call
42	specific, use `product-acceptance-verifier-maintainer` instead. If the user
43	provides only one failed run id and wants root cause, use
44	`debug-run-transcript-maintainer` first, then return here for rerun proof after
45	a fix.
46
47	## Runtime Matrix
48
49	Default required matrix:
50
51	- Codex
52	- Claude
53	- OpenCode
54	- Pi
55
56	Extend the matrix when the user names additional runtimes such as Cursor or
57	Gemini. Do not treat a Codex pass as proof for another runtime.
58
59	Read only the relevant reference files:
60
61	- `references/probe-workflow.md`: real local setup, disposable data, probe
62	script expectations, and proof levels.
63	- `references/transcript-evidence.md`: how to prove tool use and reject CLI
64	fallback from logs.
65	- `references/codex.md`: Codex MCP transcript shape and known traps.
66	- `references/claude.md`: Claude MCP transcript shape and strict config traps.
67	- `references/opencode.md`: OpenCode MCP transcript shape, provider/model
68	caveats, JSONL parser expectations, and final-text pitfalls.
69	- `references/pi.md`: Pi native extension transcript shape, schema/error
70	propagation traps, and 429/auth/provider separation.
71	- `references/reporting.md`: verdict format, mutation ledger, and pass/fail
72	language.
73
74	## Acceptance Standard
75
76	A runtime passes only when all are true:
77
78	1. The local Rudder source of truth is identified, usually
79	`GET /api/health` on `http://127.0.0.1:3100`.
80	2. A real local run was triggered for that runtime, or an existing real local
81	run was inspected with enough raw transcript/log evidence.
82	3. The transcript shows the expected Rudder MCP/native tool calls.
83	4. There is no model-visible fallback to shell, Bash, curl, or `rudder` CLI for
84	Rudder work.
85	5. Tool results are not internally failed (`isError`, structured error,
86	`rudder_cli_command_failed`, missing required argument, auth/org failure).
87	6. The terminal product effect was read back: issue status, comments, run
88	status, final text, API result, or another requested surface.
89	7. Provider/model blockers are separated from Rudder adapter/tool blockers.
90
91	If the user requires "all tools", do not only run a happy-path issue workflow.
92	Run or request a manifest-driven coverage plan: list every exposed tool, verify
93	schemas load, check runtime-managed identity is not model-provided, and execute
94	representative read/mutate/file/image/pagination/error cases. Mark unexecuted
95	tools as not covered.
96
97	## Default Probe Ladder
98
99	Use the smallest probe that answers the question, then escalate only as needed:
100
101	1. Tool availability probe: call `rudder_agent_me`, require final answer
102	to say MCP/tool path and no fallback.
103	2. Issue workflow probe: seed an issue, then require
104	`rudder_issue_context`, `rudder_issue_checkout`,
105	`rudder_issue_comment`, and `rudder_issue_done`; read back final issue
106	status and comments.
107	3. Representative matrix probe: add read/list/pagination/file/comment/image
108	examples such as `rudder_runs_errors`, `ru

People who liked this also viewed

6 picks

Type

rudder-real-runtime-verifier-maintainer

247stars30forks

View on GitHub

How to install rudder-real-runtime-verifier-maintainer?

undertone0809/rudder/rudder-real-runtime-verifier-maintainer

$npx -y skills add Undertone0809/rudder --skill rudder-real-runtime-verifier-maintainer

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# Rudder Real Runtime Verifier Maintainer
2
3	Verify Rudder agent runtime behavior on the user's real local Rudder instance.
4	This is a black-box runtime acceptance workflow, not an implementation or code
5	review workflow.
6
7	Default to Chinese when the user asks in Chinese. Put the current truth first:
8	which runtimes passed, failed, or were blocked, and what transcript evidence
9	proves it.
10
11	## Role Boundary
12
13	Default to verification and diagnosis only:
14
15	- Run real local runtime probes and inspect run transcripts, logs, API state,
16	issue state, comments, and run metadata.
17	- Create disposable orgs, agents, issues, and runs when needed for proof.
18	- Separate provider/model failure from Rudder adapter failure.
19	- Report exact blockers and smallest likely fixes.
20	- Do not edit source files, configs, git state, or product docs unless the user
21	explicitly asks for a mutation after the verification request.
22
23	If the user asks to fix the issue, hand back to the lifecycle implementation
24	route or make the smallest explicit patch, then require this skill's real
25	runtime proof again before claiming done.
26
27	## When This Skill Wins
28
29	Use this skill when the core question is whether an agent runtime actually did
30	the work through Rudder-managed tools in a real local run.
31
32	Typical prompts:
33
34	- "你所有的 agent runtime 都本地测过跑过真实环境了吗?"
35	- "OpenCode and Pi agent 你也测了吗?"
36	- "看 transcript，别让它 fallback 用 rudder cli"
37	- "排查所有 rudder tools，都试一遍"
38	- "MCP tool 报 org id/auth 问题，正常 agent 调 tool 不该传 org"
39	- "Codex/Claude/OpenCode/Pi 真实环境跑一下"
40
41	If the user asks for general product acceptance that is not runtime/tool-call
42	specific, use `product-acceptance-verifier-maintainer` instead. If the user
43	provides only one failed run id and wants root cause, use
44	`debug-run-transcript-maintainer` first, then return here for rerun proof after
45	a fix.
46
47	## Runtime Matrix
48
49	Default required matrix:
50
51	- Codex
52	- Claude
53	- OpenCode
54	- Pi
55
56	Extend the matrix when the user names additional runtimes such as Cursor or
57	Gemini. Do not treat a Codex pass as proof for another runtime.
58
59	Read only the relevant reference files:
60
61	- `references/probe-workflow.md`: real local setup, disposable data, probe
62	script expectations, and proof levels.
63	- `references/transcript-evidence.md`: how to prove tool use and reject CLI
64	fallback from logs.
65	- `references/codex.md`: Codex MCP transcript shape and known traps.
66	- `references/claude.md`: Claude MCP transcript shape and strict config traps.
67	- `references/opencode.md`: OpenCode MCP transcript shape, provider/model
68	caveats, JSONL parser expectations, and final-text pitfalls.
69	- `references/pi.md`: Pi native extension transcript shape, schema/error
70	propagation traps, and 429/auth/provider separation.
71	- `references/reporting.md`: verdict format, mutation ledger, and pass/fail
72	language.
73
74	## Acceptance Standard
75
76	A runtime passes only when all are true:
77
78	1. The local Rudder source of truth is identified, usually
79	`GET /api/health` on `http://127.0.0.1:3100`.
80	2. A real local run was triggered for that runtime, or an existing real local
81	run was inspected with enough raw transcript/log evidence.
82	3. The transcript shows the expected Rudder MCP/native tool calls.
83	4. There is no model-visible fallback to shell, Bash, curl, or `rudder` CLI for
84	Rudder work.
85	5. Tool results are not internally failed (`isError`, structured error,
86	`rudder_cli_command_failed`, missing required argument, auth/org failure).
87	6. The terminal product effect was read back: issue status, comments, run
88	status, final text, API result, or another requested surface.
89	7. Provider/model blockers are separated from Rudder adapter/tool blockers.
90
91	If the user requires "all tools", do not only run a happy-path issue workflow.
92	Run or request a manifest-driven coverage plan: list every exposed tool, verify
93	schemas load, check runtime-managed identity is not model-provided, and execute
94	representative read/mutate/file/image/pagination/error cases. Mark unexecuted
95	tools as not covered.
96
97	## Default Probe Ladder
98
99	Use the smallest probe that answers the question, then escalate only as needed:
100
101	1. Tool availability probe: call `rudder_agent_me`, require final answer
102	to say MCP/tool path and no fallback.
103	2. Issue workflow probe: seed an issue, then require
104	`rudder_issue_context`, `rudder_issue_checkout`,
105	`rudder_issue_comment`, and `rudder_issue_done`; read back final issue
106	status and comments.
107	3. Representative matrix probe: add read/list/pagination/file/comment/image
108	examples such as `rudder_runs_errors`, `ru

People who liked this also viewed

6 picks

Type