skvm-jit

Name: skvm-jit
Author: SJTU-IPADS

534stars53forks

How to install skvm-jit?

sjtu-ipads/skvm/skvm-jit

$npx -y skills add SJTU-IPADS/SkVM --skill skvm-jit

Installs into the current project.

Summary

Trigger skvm jit-optimize with post-task evidence so it generates a reviewable optimization proposal for a skill. Use after finishing a task that was driven by any skill (not just SkVM-compiled ones) when the outcome was a failure, was partial, or the skill's instructions clear

Files · 1

View on GitHub

SKILL.md

1	# SkVM JIT Optimize Trigger
2
3	When you finish a task that was driven by a skill, and the skill's own instructions appear to have caused problems, collect a short structured record of what happened and invoke `skvm jit-optimize`. The optimizer reads that evidence, proposes improvements to the skill's files, and stores them as a proposal you can inspect with `skvm proposals show <id>`.
4
5	This skill is not a human review workflow. Its job is to turn task evidence into a concrete optimization run.
6
7	This works for any skill the host harness can load — it does not need to have been produced by SkVM's compiler. `jit-optimize` only needs the skill folder on disk; it does not require SkVM-specific artifacts.
8
9	## When to trigger
10
11	Run this skill only when both of the following hold:
12
13	1. The task you just completed was driven by a skill that the harness loaded from disk (not an ad-hoc user instruction).
14	2. At least one of the following is true:
15	- the task failed or ended partial, and a clearer skill would plausibly have avoided it
16	- you observed a concrete problem in the skill's instructions: ambiguity, a missing step, an incorrect claim, a confusing ordering, or an unnecessary detour the skill forced you into
17
18	Do not run this skill when:
19
20	- the task succeeded and the skill read cleanly — silent passes are fine, the optimizer does not need "nothing to report" submissions
21	- the failure was purely user-side (typo in the prompt, missing credentials, network failure) and no change to the skill would have helped
22	- the task did not use a skill at all, or used only a trivial one-shot instruction
23	- you are running inside `skvm bench` or any other SkVM-orchestrated flow — bench owns its own optimization loop, do not double-submit
24
25	## Step 1: Locate the skill directory
26
27	The skill directory contains a `SKILL.md` file. You need the absolute path to pass as `--skill=<dir>` in Step 3.
28
29	Each agent harness installs skills in well-known locations. Read `adapter-skill-paths.md` (sibling file in this skill's directory) and look up the section matching the harness you are currently running inside — it lists the search order for Claude Code, opencode, openclaw, hermes, jiuwenclaw, and bare-agent. Probe the listed paths in order and pick the first one that contains a `SKILL.md` for the skill name you are looking for.
30
31	If none of the listed paths contains the skill, or the reference file marks your harness as "confirm with user", ask the user for the path — do not guess.
32
33	## Step 2: Prepare optimizer input
34
35	Pick one of the two formats below. Save it anywhere (e.g. a temp file); you'll pass its path as `--logs=<path>` in Step 3.
36
37	### Format A — Simple optimization report (preferred for one-off observations)
38
39	Save as `report.json`:
40
41	```json
42	{
43	"task": "<what the user asked, one or two sentences>",
44	"outcome": "pass" \| "fail" \| "partial",
45	"issues": [
46	"short description of each problem you hit",
47	"another problem"
48	],
49	"skill_feedback": "concrete suggestion for how the SKILL.md could be clearer or more correct"
50	}
51	```
52
53	Keep `issues` focused on things the skill's instructions could prevent or clarify. Do not include issues that were purely user-side (typos in the prompt, missing credentials, network failures).
54
55	The JSON key is still named `skill_feedback` because that is the current report schema consumed by `jit-optimize`; treat it as an optimization hint for the engine, not as human-directed feedback.
56
57	When `outcome` is `fail` or `partial`, the optimizer treats the issues and `skill_feedback` as failure reasons attached to a synthetic "agent-reported" criterion. When `outcome` is `pass`, the report still enters the optimizer but with no failures, letting it notice what worked well.
58
59	### Format B — Conversation log (preferred when the full turn-by-turn trace is informative)
60
61	Save as `conv-log.jsonl`, one JSON object per line:
62
63	```jsonl
64	{"type":"request","ts":"<iso8601>","text":"<user prompt>"}
65	{"type":"response","ts":"<iso8601>","text":"<your reply or summary of the step>"}
66	{"type":"tool","ts":"<iso8601>","text":"<tool call summary>","toolCalls":[...]}
67	```
68
69	Only include entries that matter for diagnosing the skill's quality. Redact secrets.
70
71	## Step 3: Run jit-optimize
72
73	```bash
74	skvm jit-optimize --detach \
75	--skill=<skill-directory> \
76	--task-source=log \
77	--logs=<path-to-report.json-or-conv-log.jsonl> \
78	--target-model=<id-the-task-ran-on> \
79	--optimizer-model=openrou

People who liked this also viewed

6 picks

Type

skvm-jit

534stars53forks

View on GitHub

How to install skvm-jit?

sjtu-ipads/skvm/skvm-jit

$npx -y skills add SJTU-IPADS/SkVM --skill skvm-jit

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# SkVM JIT Optimize Trigger
2
3	When you finish a task that was driven by a skill, and the skill's own instructions appear to have caused problems, collect a short structured record of what happened and invoke `skvm jit-optimize`. The optimizer reads that evidence, proposes improvements to the skill's files, and stores them as a proposal you can inspect with `skvm proposals show <id>`.
4
5	This skill is not a human review workflow. Its job is to turn task evidence into a concrete optimization run.
6
7	This works for any skill the host harness can load — it does not need to have been produced by SkVM's compiler. `jit-optimize` only needs the skill folder on disk; it does not require SkVM-specific artifacts.
8
9	## When to trigger
10
11	Run this skill only when both of the following hold:
12
13	1. The task you just completed was driven by a skill that the harness loaded from disk (not an ad-hoc user instruction).
14	2. At least one of the following is true:
15	- the task failed or ended partial, and a clearer skill would plausibly have avoided it
16	- you observed a concrete problem in the skill's instructions: ambiguity, a missing step, an incorrect claim, a confusing ordering, or an unnecessary detour the skill forced you into
17
18	Do not run this skill when:
19
20	- the task succeeded and the skill read cleanly — silent passes are fine, the optimizer does not need "nothing to report" submissions
21	- the failure was purely user-side (typo in the prompt, missing credentials, network failure) and no change to the skill would have helped
22	- the task did not use a skill at all, or used only a trivial one-shot instruction
23	- you are running inside `skvm bench` or any other SkVM-orchestrated flow — bench owns its own optimization loop, do not double-submit
24
25	## Step 1: Locate the skill directory
26
27	The skill directory contains a `SKILL.md` file. You need the absolute path to pass as `--skill=<dir>` in Step 3.
28
29	Each agent harness installs skills in well-known locations. Read `adapter-skill-paths.md` (sibling file in this skill's directory) and look up the section matching the harness you are currently running inside — it lists the search order for Claude Code, opencode, openclaw, hermes, jiuwenclaw, and bare-agent. Probe the listed paths in order and pick the first one that contains a `SKILL.md` for the skill name you are looking for.
30
31	If none of the listed paths contains the skill, or the reference file marks your harness as "confirm with user", ask the user for the path — do not guess.
32
33	## Step 2: Prepare optimizer input
34
35	Pick one of the two formats below. Save it anywhere (e.g. a temp file); you'll pass its path as `--logs=<path>` in Step 3.
36
37	### Format A — Simple optimization report (preferred for one-off observations)
38
39	Save as `report.json`:
40
41	```json
42	{
43	"task": "<what the user asked, one or two sentences>",
44	"outcome": "pass" \| "fail" \| "partial",
45	"issues": [
46	"short description of each problem you hit",
47	"another problem"
48	],
49	"skill_feedback": "concrete suggestion for how the SKILL.md could be clearer or more correct"
50	}
51	```
52
53	Keep `issues` focused on things the skill's instructions could prevent or clarify. Do not include issues that were purely user-side (typos in the prompt, missing credentials, network failures).
54
55	The JSON key is still named `skill_feedback` because that is the current report schema consumed by `jit-optimize`; treat it as an optimization hint for the engine, not as human-directed feedback.
56
57	When `outcome` is `fail` or `partial`, the optimizer treats the issues and `skill_feedback` as failure reasons attached to a synthetic "agent-reported" criterion. When `outcome` is `pass`, the report still enters the optimizer but with no failures, letting it notice what worked well.
58
59	### Format B — Conversation log (preferred when the full turn-by-turn trace is informative)
60
61	Save as `conv-log.jsonl`, one JSON object per line:
62
63	```jsonl
64	{"type":"request","ts":"<iso8601>","text":"<user prompt>"}
65	{"type":"response","ts":"<iso8601>","text":"<your reply or summary of the step>"}
66	{"type":"tool","ts":"<iso8601>","text":"<tool call summary>","toolCalls":[...]}
67	```
68
69	Only include entries that matter for diagnosing the skill's quality. Redact secrets.
70
71	## Step 3: Run jit-optimize
72
73	```bash
74	skvm jit-optimize --detach \
75	--skill=<skill-directory> \
76	--task-source=log \
77	--logs=<path-to-report.json-or-conv-log.jsonl> \
78	--target-model=<id-the-task-ran-on> \
79	--optimizer-model=openrou

People who liked this also viewed

6 picks

Type