ciunkos 0.3.3

Install in Claude Code:

/plugin marketplace add Ciunkos/skills
/plugin install ciunkos@ciunkos

Or upload a .skill file below to claude.ai or Claude Desktop.

business-idea-validator 0.1.0

First public release.

• Ten rubrics with scoring anchors, a fixed verdict rule (Build, Validate first, Pivot, Stop) and a critic pass.
• Markdown and JSON report in English or Polish, with the riskiest assumption, up to three experiments and open questions.
• scripts/check.mjs validates a report against the schemas, computes the verdict and renders Markdown.
• Report footer is a plain AI disclaimer: generated with AI, can make mistakes, scores are a starting point.

przemo 0.2.3

• references/habits.md: how he answers project questions, with invented examples. Root cause over patch, delete first, evidence first, "problem or guess", question the numbers, labeled guesses, YAGNI, restate the ask and "Right?", conditional yes to a first iteration, business calls go to the owner.
• The key habits are also in SKILL.md, so they apply even when references are not read.

ciunkos 0.3.2

Install in Claude Code:

/plugin marketplace add Ciunkos/skills
/plugin install ciunkos@ciunkos

Or upload a .skill file below to claude.ai or Claude Desktop.

business-idea-validator 0.1.0

First public release.

• Ten rubrics with scoring anchors, a fixed verdict rule (Build, Validate first, Pivot, Stop) and a critic pass.
• Markdown and JSON report in English or Polish, with the riskiest assumption, up to three experiments and open questions.
• scripts/check.mjs validates a report against the schemas, computes the verdict and renders Markdown.
• Report footer is a plain AI disclaimer: generated with AI, can make mistakes, scores are a starting point.

przemo 0.2.2

• New habits: merge small safe changes and review after, no conditional UI, remove dead flags after checking why they exist, ask for proof of claimed deadlines, polish before features, no support for outdated browsers.
• Przemo never repeats himself: no template openers or closers, the late-hour line is rare and new each time. The only sure thing is that he answers.
• Voice: the technical catch in parentheses, open self-correction, concrete numbers in suggestions, warm side notes.

ciunkos 0.3.1

Install in Claude Code:

/plugin marketplace add Ciunkos/skills
/plugin install ciunkos@ciunkos

Or upload a .skill file below to claude.ai or Claude Desktop.

business-idea-validator 0.1.0

First public release.

• Ten rubrics with scoring anchors, a fixed verdict rule (Build, Validate first, Pivot, Stop) and a critic pass.
• Markdown and JSON report in English or Polish, with the riskiest assumption, up to three experiments and open questions.
• scripts/check.mjs validates a report against the schemas, computes the verdict and renders Markdown.
• Report footer is a plain AI disclaimer: generated with AI, can make mistakes, scores are a starting point.

przemo 0.2.1

• Consult mode is short by default and answers "how should it work" as product behavior first.
• Cheap, sensible options get a yes instead of a separate ticket; existing data stays untouched and an empty value means no change.
• "Moje 50¢" at most once per conversation; most replies end on the substance.

ciunkos 0.3.0

Install in Claude Code:

/plugin marketplace add Ciunkos/skills
/plugin install ciunkos@ciunkos

Or upload a .skill file below to claude.ai or Claude Desktop.

business-idea-validator 0.1.0

First public release.

• Ten rubrics with scoring anchors, a fixed verdict rule (Build, Validate first, Pivot, Stop) and a critic pass.
• Markdown and JSON report in English or Polish, with the riskiest assumption, up to three experiments and open questions.
• scripts/check.mjs validates a report against the schemas, computes the verdict and renders Markdown.
• Report footer is a plain AI disclaimer: generated with AI, can make mistakes, scores are a starting point.

przemo 0.2.0

• Review mode: product first, deletion first, root cause over hacks, security checklist, a one-line verdict and Blokery / Warto / Nity with file:line and a fix for each.
• Consult mode for hard calls and stuck bugs: one recommendation, the deciding reason, the alternative, the first step.
• Work mode reads references/craft.md: how Przemo writes code and runs agents, with a second pass when a fix looks like a hack.
• Public facts match the hire page. Self-appointed joke titles added.
• Calls the real one at most once per conversation, only when production is down or the problem is bigger than a session.

ciunkos 0.2.0

Install in Claude Code:

/plugin marketplace add Ciunkos/skills
/plugin install ciunkos@ciunkos

Or upload a .skill file below to claude.ai or Claude Desktop.

business-idea-validator 0.1.0

First public release.

• Ten rubrics with scoring anchors, a fixed verdict rule (Build, Validate first, Pivot, Stop) and a critic pass.
• Markdown and JSON report in English or Polish, with the riskiest assumption, up to three experiments and open questions.
• scripts/check.mjs validates a report against the schemas, computes the verdict and renders Markdown.
• Report footer is a plain AI disclaimer: generated with AI, can make mistakes, scores are a starting point.

przemo 0.1.0

First public release.

• Talk mode: short Slack-style replies in Polish (English when you write in English), one signature move per reply, no repeats.
• Work mode: does the task with a success criterion, root cause, a reproducing test and a self-check before "done".
• Reads the local time once and has one line about late-night deploys.
• Humor as on Flack (ciunkos.com/przemo), with black humor and politics on request and hard limits on targets.

przemo 0.2.3

• references/habits.md: how he answers project questions, with invented examples. Root cause over patch, delete first, evidence first, "problem or guess", question the numbers, labeled guesses, YAGNI, restate the ask and "Right?", conditional yes to a first iteration, business calls go to the owner.
• The key habits are also in SKILL.md, so they apply even when references are not read.

przemo 0.2.2

• New habits: merge small safe changes and review after, no conditional UI, remove dead flags after checking why they exist, ask for proof of claimed deadlines, polish before features, no support for outdated browsers.
• Przemo never repeats himself: no template openers or closers, the late-hour line is rare and new each time. The only sure thing is that he answers.
• Voice: the technical catch in parentheses, open self-correction, concrete numbers in suggestions, warm side notes.

przemo 0.2.1

• Consult mode is short by default and answers "how should it work" as product behavior first.
• Cheap, sensible options get a yes instead of a separate ticket; existing data stays untouched and an empty value means no change.
• "Moje 50¢" at most once per conversation; most replies end on the substance.

przemo 0.2.0

• Review mode: product first, deletion first, root cause over hacks, security checklist, a one-line verdict and Blokery / Warto / Nity with file:line and a fix for each.
• Consult mode for hard calls and stuck bugs: one recommendation, the deciding reason, the alternative, the first step.
• Work mode reads references/craft.md: how Przemo writes code and runs agents, with a second pass when a fix looks like a hack.
• Public facts match the hire page. Self-appointed joke titles added.
• Calls the real one at most once per conversation, only when production is down or the problem is bigger than a session.

przemo 0.1.0

First public release.

• Talk mode: short Slack-style replies in Polish (English when you write in English), one signature move per reply, no repeats.
• Work mode: does the task with a success criterion, root cause, a reproducing test and a self-check before "done".
• Reads the local time once and has one line about late-night deploys.
• Humor as on Flack (ciunkos.com/przemo), with black humor and politics on request and hard limits on targets.