synthesize-transcript

Name: synthesize-transcript
Author: vercel-labs

202stars12forks

How to install synthesize-transcript?

vercel-labs/next-browser/synthesize-transcript

$npx -y skills add vercel-labs/next-browser --skill synthesize-transcript

Installs into the current project.

Summary

Analyze an agent transcript for generalizable learnings about next-browser CLI usage, then propose SKILL.md updates, feature requests, and bug fixes.

Files · 1

View on GitHub

SKILL.md

1	# synthesize-transcript
2
3	Read the transcript at:
4
5	$ARGUMENTS
6
7	Then analyze how the agent used `next-browser` commands. Your goals are to:
8	1. Find generalizable learnings — things that would help any agent using
9	this tool — and propose targeted additions to `SKILL.md`.
10	2. Surface feature requests and bug fixes for the CLI itself — gaps in
11	functionality, incorrect behavior, or misleading output that should be
12	fixed in code.
13
14	## Process
15
16	1. Read the transcript in chunks (it's large). Focus on assistant
17	messages and tool calls/results involving `next-browser` commands.
18
19	2. Identify friction and patterns. Look for:
20	- Commands that failed or produced unexpected results
21	- Misunderstandings about how a command works
22	- Workflows the agent discovered through trial and error
23	- Repeated mistakes that guidance would prevent
24
25	3. Filter ruthlessly for overfitting. For each candidate learning, ask:
26	- Would a different agent on a different project hit this same issue?
27	- Is this a property of the tool or a property of this debugging session?
28	- Is this already obvious from the command's existing docs?
29	- Is this prescribing a specific workflow vs documenting tool behavior?
30
31	Discard anything that is:
32	- A workflow pattern specific to one task (e.g., "copy dirs for before/after")
33	- Advice an agent could derive from reading existing docs
34	- Specific to a particular project, page, or debugging scenario
35
36	Keep as SKILL.md learnings:
37	- Genuine tool constraints any agent would hit (e.g., eval quirks, Playwright behavior)
38	- Non-obvious failure modes with clear mitigations
39	- Command interactions that aren't documented
40
41	Separate out as feature requests or bug fixes (not SKILL.md changes):
42	- Workarounds for bugs that should be fixed in code
43	- Missing functionality the agent needed and had to hack around
44	- Misleading output or docs that don't match actual behavior
45
46	4. Present your findings. Organize into two sections:
47
48	SKILL.md learnings — for each, show:
49	- What happened in the transcript (brief)
50	- The proposed SKILL.md addition (exact text)
51	- Why it's generalizable (one sentence)
52
53	Feature requests / bug fixes — for each, show:
54	- What happened in the transcript (brief)
55	- What the CLI should do instead (proposed behavior)
56	- Whether it's a bug fix (current behavior is wrong) or a feature
57	request (new capability needed)
58
59	Ask the user to approve or reject each item before acting.
60
61	5. Consider new Scenarios. Beyond command-level learnings, check
62	whether the transcript reveals a scenario not covered in the existing
63	`## Scenarios` section of SKILL.md. A scenario earns its place only if
64	it requires domain knowledge an agent wouldn't derive on its own —
65	non-obvious mental models, ordering constraints, or decision frameworks
66	that go beyond "use these commands and compare results."
67
68	Ask yourself: could an agent figure out this workflow just by reading
69	the existing command docs and applying basic debugging instincts? If
70	yes, it's not a scenario — it's just competent tool use. If no — if it
71	requires understanding something about React, Next.js, or the PPR model
72	that isn't in the CLI docs — then it's a candidate.
73
74	Present scenario candidates the same way as command learnings: evidence
75	from the transcript, proposed text, and why an agent couldn't get there
76	alone. Ask for approval before adding.
77
78	6. Apply approved changes.
79	- SKILL.md learnings → add inline to the relevant command section in
80	the repo-root `SKILL.md`. Scenarios → `## Scenarios` section.
81	- Feature requests / bug fixes → we own this package, so implement
82	the fix or feature directly in the codebase. For non-trivial
83	changes, plan the implementation and confirm with the user before
84	writing code. Include the transcript evidence in commit messages
85	or PR descriptions for context.
86
87	## Anti-patterns
88
89	- Don't add "tips" or "best practices" sections. Guidance belongs next to
90	the command it's about.
91	- Don't add workflow recipes ("first do X, then Y, then Z"). Document tool
92	behavior, not agent strategy.
93	- Don't inflate existing docs with caveats that rarely apply.
94	- When in doubt, leave it out. A wrong or noisy addition is worse than a
95	missing one.

People who liked this also viewed

6 picks

Type

ai-video-generationinference.sh Agent skills for using our API to give your agents access to hundreds of apps and other agentsSkillsJul 2026292k642
imagegen-frontend-webElite frontend image-direction skill for generating premium, conversion-aware website design references. CRITICAL OUTPUT RULE — generate ONE separate horizontal image FOR EVERY section. A landing pageSkillsJul 2026165k66k
hyperframes-mediaWrite HTML. Render video. Built for agents.SkillsJul 2026145k37k
media-useAgent Media OS, the single skill for every media need in a HyperFrames project. Resolve BGM, SFX, image, icon, brand logo, voice, color grade, or LUT into a frozen local file or paste-ready block + leSkillsJul 2026144k37k
general-videoAuthor or edit a custom HyperFrames composition when no specialized workflow fits, or when BRIEF.md sets flow: companion. Use for longer or multi-scene pieces, brand and sizzle reels, montages, staticSkillsJul 2026124k37k
product-launch-videoTurn a product or marketing URL, pasted script, or brief into a product launch / promo video — SaaS promos, feature reveals, product demos, app and company launches. Use when the user wants to market,SkillsJul 2026121k37k

synthesize-transcript

202stars12forks

View on GitHub

How to install synthesize-transcript?

vercel-labs/next-browser/synthesize-transcript

$npx -y skills add vercel-labs/next-browser --skill synthesize-transcript

Installs into the current project.

Summary

Analyze an agent transcript for generalizable learnings about next-browser CLI usage, then propose SKILL.md updates, feature requests, and bug fixes.

Files · 1

View on GitHub

SKILL.md

1	# synthesize-transcript
2
3	Read the transcript at:
4
5	$ARGUMENTS
6
7	Then analyze how the agent used `next-browser` commands. Your goals are to:
8	1. Find generalizable learnings — things that would help any agent using
9	this tool — and propose targeted additions to `SKILL.md`.
10	2. Surface feature requests and bug fixes for the CLI itself — gaps in
11	functionality, incorrect behavior, or misleading output that should be
12	fixed in code.
13
14	## Process
15
16	1. Read the transcript in chunks (it's large). Focus on assistant
17	messages and tool calls/results involving `next-browser` commands.
18
19	2. Identify friction and patterns. Look for:
20	- Commands that failed or produced unexpected results
21	- Misunderstandings about how a command works
22	- Workflows the agent discovered through trial and error
23	- Repeated mistakes that guidance would prevent
24
25	3. Filter ruthlessly for overfitting. For each candidate learning, ask:
26	- Would a different agent on a different project hit this same issue?
27	- Is this a property of the tool or a property of this debugging session?
28	- Is this already obvious from the command's existing docs?
29	- Is this prescribing a specific workflow vs documenting tool behavior?
30
31	Discard anything that is:
32	- A workflow pattern specific to one task (e.g., "copy dirs for before/after")
33	- Advice an agent could derive from reading existing docs
34	- Specific to a particular project, page, or debugging scenario
35
36	Keep as SKILL.md learnings:
37	- Genuine tool constraints any agent would hit (e.g., eval quirks, Playwright behavior)
38	- Non-obvious failure modes with clear mitigations
39	- Command interactions that aren't documented
40
41	Separate out as feature requests or bug fixes (not SKILL.md changes):
42	- Workarounds for bugs that should be fixed in code
43	- Missing functionality the agent needed and had to hack around
44	- Misleading output or docs that don't match actual behavior
45
46	4. Present your findings. Organize into two sections:
47
48	SKILL.md learnings — for each, show:
49	- What happened in the transcript (brief)
50	- The proposed SKILL.md addition (exact text)
51	- Why it's generalizable (one sentence)
52
53	Feature requests / bug fixes — for each, show:
54	- What happened in the transcript (brief)
55	- What the CLI should do instead (proposed behavior)
56	- Whether it's a bug fix (current behavior is wrong) or a feature
57	request (new capability needed)
58
59	Ask the user to approve or reject each item before acting.
60
61	5. Consider new Scenarios. Beyond command-level learnings, check
62	whether the transcript reveals a scenario not covered in the existing
63	`## Scenarios` section of SKILL.md. A scenario earns its place only if
64	it requires domain knowledge an agent wouldn't derive on its own —
65	non-obvious mental models, ordering constraints, or decision frameworks
66	that go beyond "use these commands and compare results."
67
68	Ask yourself: could an agent figure out this workflow just by reading
69	the existing command docs and applying basic debugging instincts? If
70	yes, it's not a scenario — it's just competent tool use. If no — if it
71	requires understanding something about React, Next.js, or the PPR model
72	that isn't in the CLI docs — then it's a candidate.
73
74	Present scenario candidates the same way as command learnings: evidence
75	from the transcript, proposed text, and why an agent couldn't get there
76	alone. Ask for approval before adding.
77
78	6. Apply approved changes.
79	- SKILL.md learnings → add inline to the relevant command section in
80	the repo-root `SKILL.md`. Scenarios → `## Scenarios` section.
81	- Feature requests / bug fixes → we own this package, so implement
82	the fix or feature directly in the codebase. For non-trivial
83	changes, plan the implementation and confirm with the user before
84	writing code. Include the transcript evidence in commit messages
85	or PR descriptions for context.
86
87	## Anti-patterns
88
89	- Don't add "tips" or "best practices" sections. Guidance belongs next to
90	the command it's about.
91	- Don't add workflow recipes ("first do X, then Y, then Z"). Document tool
92	behavior, not agent strategy.
93	- Don't inflate existing docs with caveats that rarely apply.
94	- When in doubt, leave it out. A wrong or noisy addition is worse than a
95	missing one.

People who liked this also viewed

6 picks

Type