skvm-general

Name: skvm-general
Author: SJTU-IPADS

534stars53forks

How to install skvm-general?

sjtu-ipads/skvm/skvm-general

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

Installs into the current project.

Summary

Drive the skvm CLI on behalf of a user to profile models, AOT-compile skills, run skill-assisted tasks, run benchmarks, and manage compiled proposals. Trigger when the user asks to "profile", "aot-compile", "bench", "run a single ad-hoc task with a skill", or asks about skvm prop

Files · 1

View on GitHub

SKILL.md

1	# SkVM General Usage
2
3	You are driving `skvm`, a CLI that AOT-compiles and runs LLM agent skills across heterogeneous models. Use this skill when the user wants to use skvm — profile a model, AOT-compile a skill, run a task with a skill, run a benchmark, or manage optimization proposals. Do not invent flags — every example below uses the real flag set from the installed `skvm` binary.
4
5	## Step 1: Prerequisite self-check
6
7	Split the check in two — the binary must always be present, but the API key is only required for commands that call an LLM.
8
9	Always required — skvm is on PATH:
10
11	```bash
12	skvm --help >/dev/null 2>&1 \|\| { echo "skvm not installed — tell the user to run: curl -fsSL https://skillvm.ai/install.sh \| sh"; exit 1; }
13	```
14
15	Required only before LLM-calling commands — `profile` (without `--list`), `aot-compile`, `pipeline`, `run`, `bench`, `jit-optimize`. Local filesystem commands (`profile --list`, `proposals list\|show\|reject`, `logs`, `clean-jit`) do not need the API key — run them even if the key is unset.
16
17	```bash
18	# Before running profile/aot-compile/run/bench/jit-optimize:
19	test -n "${OPENROUTER_API_KEY:-}" \|\| { echo "OPENROUTER_API_KEY is not set — ask the user for their key"; exit 1; }
20	```
21
22	If a required prerequisite is missing, stop and tell the user what is missing. Do not install anything yourself.
23
24	## Step 2: Profile a model
25
26	A profile (TCP — Target Capability Profile) records what an LLM can do across 26 primitive capabilities. It is the input for AOT compilation and is cached so subsequent compile calls reuse it.
27
28	```bash
29	skvm profile --model=<id> # profile one model
30	skvm profile --model=<id1>,<id2> --concurrency=4 # profile several in parallel
31	skvm profile --model=<id> --adapter=opencode # non-default adapter
32	skvm profile --model=<id> --force # ignore cache, re-run
33	skvm profile --list # list cached profiles
34	```
35
36	Notes:
37	- Default adapter is `bare-agent`. Other valid adapters: `opencode`, `openclaw`, `hermes`, `jiuwenclaw`.
38	- Cache lives at `$SKVM_PROFILES_DIR` (default `.skvm/profiles/`).
39	- Profiling is expensive — confirm with the user before running on several models, and prefer `--concurrency` over sequential runs.
40
41	## Step 3: AOT-compile a skill
42
43	AOT compilation rewrites a skill's `SKILL.md` (and optionally bundle files) so it fits a specific target model's capability profile. The three-pass AOT compiler runs by default.
44
45	```bash
46	skvm aot-compile --skill=<path> --model=<id> # all three passes
47	skvm aot-compile --skill=<path> --model=<id> --pass=1,2,3 # explicit
48	skvm aot-compile --skill=<path> --model=<id> --pass=1 # only pass 1 (SCR + gap analysis)
49	skvm aot-compile --skill=<path> --model=<id> --dry-run # no write
50	skvm pipeline --skill=<path> --model=<id> # profile-if-needed → aot-compile
51	```
52
53	Pass semantics:
54	- `--pass=1` — SCR extraction, gap analysis, capability substitution/compensation
55	- `--pass=2` — dependency manifest + env-binding script generation
56	- `--pass=3` — workflow decomposition + DAG parallelism extraction
57
58	Compiled variants land under the proposals tree (`proposals/aot-compile/...`). Multiple passes can be combined in any subset: `--pass=1,3` runs passes 1 and 3, skipping 2.
59
60	## Step 4: Run a single task with a skill
61
62	For ad-hoc debugging of one skill on one task:
63
64	```bash
65	skvm run --task=<path/to/task.json> --model=<id> # no skill
66	skvm run --task=<path> --model=<id> --skill=<path/to/SKILL.md> # with skill
67	skvm run --task=<path> --model=<id> --adapter=opencode --verbose # explicit adapter + debug
68	```
69
70	Use this to reproduce a single failing task or validate a skill edit. Do not use it for benchmarking — use `skvm bench` instead.
71
72	## Step 5: Bench a skill
73
74	Benchmarking runs a skill across many tasks and condition variants. It can get expensive fast — always confirm with the user before running across many models or tasks, and use `--concurrency` for parallelism.
75
76	```bash
77	skvm bench --model=<id> # all conditions, all tasks
78	skvm bench --model=<id> --conditions=original,aot-compiled # baseline + compiled
79	skvm bench --model=<id> --conditions=jit-optimized # use latest jit-optimize best round
80	skvm bench --model=<id> --conditions=jit-boost --jit-runs=5 # 5 warmup runs for solidification
81	skvm bench --model=<id1>,<id2> --conc

People who liked this also viewed

6 picks

Type

skvm-general

534stars53forks

View on GitHub

How to install skvm-general?

sjtu-ipads/skvm/skvm-general

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

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# SkVM General Usage
2
3	You are driving `skvm`, a CLI that AOT-compiles and runs LLM agent skills across heterogeneous models. Use this skill when the user wants to use skvm — profile a model, AOT-compile a skill, run a task with a skill, run a benchmark, or manage optimization proposals. Do not invent flags — every example below uses the real flag set from the installed `skvm` binary.
4
5	## Step 1: Prerequisite self-check
6
7	Split the check in two — the binary must always be present, but the API key is only required for commands that call an LLM.
8
9	Always required — skvm is on PATH:
10
11	```bash
12	skvm --help >/dev/null 2>&1 \|\| { echo "skvm not installed — tell the user to run: curl -fsSL https://skillvm.ai/install.sh \| sh"; exit 1; }
13	```
14
15	Required only before LLM-calling commands — `profile` (without `--list`), `aot-compile`, `pipeline`, `run`, `bench`, `jit-optimize`. Local filesystem commands (`profile --list`, `proposals list\|show\|reject`, `logs`, `clean-jit`) do not need the API key — run them even if the key is unset.
16
17	```bash
18	# Before running profile/aot-compile/run/bench/jit-optimize:
19	test -n "${OPENROUTER_API_KEY:-}" \|\| { echo "OPENROUTER_API_KEY is not set — ask the user for their key"; exit 1; }
20	```
21
22	If a required prerequisite is missing, stop and tell the user what is missing. Do not install anything yourself.
23
24	## Step 2: Profile a model
25
26	A profile (TCP — Target Capability Profile) records what an LLM can do across 26 primitive capabilities. It is the input for AOT compilation and is cached so subsequent compile calls reuse it.
27
28	```bash
29	skvm profile --model=<id> # profile one model
30	skvm profile --model=<id1>,<id2> --concurrency=4 # profile several in parallel
31	skvm profile --model=<id> --adapter=opencode # non-default adapter
32	skvm profile --model=<id> --force # ignore cache, re-run
33	skvm profile --list # list cached profiles
34	```
35
36	Notes:
37	- Default adapter is `bare-agent`. Other valid adapters: `opencode`, `openclaw`, `hermes`, `jiuwenclaw`.
38	- Cache lives at `$SKVM_PROFILES_DIR` (default `.skvm/profiles/`).
39	- Profiling is expensive — confirm with the user before running on several models, and prefer `--concurrency` over sequential runs.
40
41	## Step 3: AOT-compile a skill
42
43	AOT compilation rewrites a skill's `SKILL.md` (and optionally bundle files) so it fits a specific target model's capability profile. The three-pass AOT compiler runs by default.
44
45	```bash
46	skvm aot-compile --skill=<path> --model=<id> # all three passes
47	skvm aot-compile --skill=<path> --model=<id> --pass=1,2,3 # explicit
48	skvm aot-compile --skill=<path> --model=<id> --pass=1 # only pass 1 (SCR + gap analysis)
49	skvm aot-compile --skill=<path> --model=<id> --dry-run # no write
50	skvm pipeline --skill=<path> --model=<id> # profile-if-needed → aot-compile
51	```
52
53	Pass semantics:
54	- `--pass=1` — SCR extraction, gap analysis, capability substitution/compensation
55	- `--pass=2` — dependency manifest + env-binding script generation
56	- `--pass=3` — workflow decomposition + DAG parallelism extraction
57
58	Compiled variants land under the proposals tree (`proposals/aot-compile/...`). Multiple passes can be combined in any subset: `--pass=1,3` runs passes 1 and 3, skipping 2.
59
60	## Step 4: Run a single task with a skill
61
62	For ad-hoc debugging of one skill on one task:
63
64	```bash
65	skvm run --task=<path/to/task.json> --model=<id> # no skill
66	skvm run --task=<path> --model=<id> --skill=<path/to/SKILL.md> # with skill
67	skvm run --task=<path> --model=<id> --adapter=opencode --verbose # explicit adapter + debug
68	```
69
70	Use this to reproduce a single failing task or validate a skill edit. Do not use it for benchmarking — use `skvm bench` instead.
71
72	## Step 5: Bench a skill
73
74	Benchmarking runs a skill across many tasks and condition variants. It can get expensive fast — always confirm with the user before running across many models or tasks, and use `--concurrency` for parallelism.
75
76	```bash
77	skvm bench --model=<id> # all conditions, all tasks
78	skvm bench --model=<id> --conditions=original,aot-compiled # baseline + compiled
79	skvm bench --model=<id> --conditions=jit-optimized # use latest jit-optimize best round
80	skvm bench --model=<id> --conditions=jit-boost --jit-runs=5 # 5 warmup runs for solidification
81	skvm bench --model=<id1>,<id2> --conc

People who liked this also viewed

6 picks

Type