<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:dc="http://purl.org/dc/elements/1.1/">
  <channel>
    <title>DEV Community: Mátyás Kuti-Kreszács</title>
    <description>The latest articles on DEV Community by Mátyás Kuti-Kreszács (@thec0dewriter).</description>
    <link>https://dev.to/thec0dewriter</link>
    <image>
      <url>https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F3740463%2F938c438d-831e-43d8-9f48-71decc7e2de6.png</url>
      <title>DEV Community: Mátyás Kuti-Kreszács</title>
      <link>https://dev.to/thec0dewriter</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://dev.to/feed/thec0dewriter"/>
    <language>en</language>
    <item>
      <title>Catalog-bound Decide+Act: why free tool-calling felt wrong for SysOps writes</title>
      <dc:creator>Mátyás Kuti-Kreszács</dc:creator>
      <pubDate>Tue, 17 Nov 2026 19:38:28 +0000</pubDate>
      <link>https://dev.to/thec0dewriter/catalog-bound-decideact-why-free-tool-calling-felt-wrong-for-sysops-writes-118b</link>
      <guid>https://dev.to/thec0dewriter/catalog-bound-decideact-why-free-tool-calling-felt-wrong-for-sysops-writes-118b</guid>
      <description>&lt;p&gt;&lt;em&gt;This is a submission for the &lt;a href="https://dev.to/mlh-hackathon"&gt;MLH x DEV Writing Challenge&lt;/a&gt;&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;I spent &lt;a href="https://ghw.mlh.com/events/builders-week" rel="noopener noreferrer"&gt;MLH Global Hack Week: Builders Week (Nov 6–12, 2026)&lt;/a&gt; fighting a pattern that still shows up in “AI Ops” demos: an LLM that &lt;em&gt;talks&lt;/em&gt; about fixing a host, then invents the shell command that would do it.&lt;/p&gt;

&lt;p&gt;This is a builder story — process and trust boundaries, not a preprint.&lt;/p&gt;

&lt;h2&gt;
  
  
  What I Built
&lt;/h2&gt;

&lt;p&gt;&lt;strong&gt;Escalane&lt;/strong&gt; is a catalog-bound &lt;strong&gt;Decide+Act&lt;/strong&gt; harness for SysOps. The model chooses &lt;em&gt;what&lt;/em&gt; kind of fix fits; the harness is the only thing that &lt;em&gt;runs&lt;/em&gt;. Code + CLI: &lt;a href="https://github.com/escalane-lab/escalane" rel="noopener noreferrer"&gt;&lt;code&gt;escalane-lab/escalane&lt;/code&gt;&lt;/a&gt;.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;System One&lt;/strong&gt; is the decision backend — it answers typed questions (which playbook, safe to mutate, how wide the blast radius is) over a scrubbed state blob. It does not get a free tool belt.&lt;/p&gt;

&lt;p&gt;Why not “Agent = LLM + tools”? In SysOps, a “tool” is often a shell-shaped surface. Free tool-calling collapses three jobs that should not share a brain:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Judgment&lt;/strong&gt; — which remediation class fits?&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Authority&lt;/strong&gt; — is it safe to mutate &lt;em&gt;this&lt;/em&gt; host &lt;em&gt;now&lt;/em&gt;?&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Execution&lt;/strong&gt; — exact shell args, locks, dry-run, verify&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;When the model authors commands, prompt-injected telemetry becomes a remote-code path. I wanted a closed action space: pick a playbook id (or escalate), answer safety questions, and let harness-owned &lt;strong&gt;catalog templates&lt;/strong&gt; fill shell args from &lt;strong&gt;state-bound&lt;/strong&gt; params — never from model free text.&lt;/p&gt;

&lt;p&gt;Closing the &lt;em&gt;verbs&lt;/em&gt; is not enough. If a unit name or container id in the state blob is attacker-influenced, that &lt;strong&gt;noun&lt;/strong&gt; can still reach argv through the harness. “Came from state” is provenance, not authority. So each template slot can declare a type (&lt;code&gt;x-escalane-slot&lt;/code&gt;) and an &lt;strong&gt;allowlist&lt;/strong&gt; (&lt;code&gt;enum&lt;/code&gt; / catalog &lt;code&gt;allowlists&lt;/code&gt;), checked at bind &lt;strong&gt;and&lt;/strong&gt; again at Act time. Ingest still rejects command-like fields; that stays necessary and insufficient.&lt;/p&gt;

&lt;p&gt;One simple loop:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Ingest&lt;/strong&gt; — symptoms only; reject command-like fields&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Decide&lt;/strong&gt; — System One returns typed answers (choice / noul / blast score)&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Gates&lt;/strong&gt; — ordered checks; fail closed when &lt;strong&gt;external&lt;/strong&gt; harness authority is missing (precision, lock, allowlist, precheck) — not merely when the model sounds unsure&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Catalog act&lt;/strong&gt; — template argv only after allowlist re-check, or escalate / ask a human&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;Research default is dry-run (no host mutation). Writes need dual consent: CLI &lt;code&gt;--allow-writes&lt;/code&gt; &lt;em&gt;and&lt;/em&gt; &lt;code&gt;ALLOW_WRITE_OPERATIONS=true&lt;/code&gt;. Fail-closed &lt;code&gt;--require-live&lt;/code&gt; refuses silent mock fallback when you asked for a live backend.&lt;/p&gt;

&lt;p&gt;Lesson that stuck: &lt;strong&gt;harness &amp;gt; model swap&lt;/strong&gt;. Changing the decision model does not fix unsafe shell-arg authorship. Grammar ≠ semantic safety. Provenance ≠ allowlist. Confidence ≠ authority.&lt;/p&gt;

&lt;h2&gt;
  
  
  Demo
&lt;/h2&gt;

&lt;p&gt;Docs site: &lt;a href="https://escalane-lab.github.io/escalane-docs/" rel="noopener noreferrer"&gt;escalane-lab.github.io/escalane-docs&lt;/a&gt;. The usable path is a local CLI walkthrough (screenshot below). Dry-run is the default: nothing mutates a host unless you intentionally unlock writes. The model never writes shell args — only catalog templates do.&lt;/p&gt;

&lt;p&gt;&lt;code&gt;--gate-config&lt;/code&gt; picks how strict the ordered safety checks are. Gates &lt;strong&gt;A0–A3&lt;/strong&gt; run in order; each can veto before anything runs:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;A0&lt;/strong&gt; — is the chosen playbook in the catalog? (harness membership)&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;A1&lt;/strong&gt; — is noul / confidence high enough? (&lt;strong&gt;model self-report&lt;/strong&gt; — useful filter, not proof; a steered choice has no reason to look unconfident)&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;A2&lt;/strong&gt; — is blast radius acceptable against harness thresholds?&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;A3&lt;/strong&gt; — is fitted &lt;strong&gt;per-label precision&lt;/strong&gt; (harness empirical table) good enough? If not → &lt;strong&gt;human review&lt;/strong&gt; (HITL)&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;&lt;strong&gt;A3&lt;/strong&gt; is the research posture I actually demo for mutate paths. On this &lt;strong&gt;live&lt;/strong&gt; Ollama run with System One &lt;code&gt;d3:4b&lt;/code&gt;, it still asked for a human, because fitted label precision was conservative — even though the model picked a valid catalog playbook and reported high noul.&lt;/p&gt;

&lt;p&gt;Live decide (no host mutation; fail-closed if the backend is down):&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;uv sync
uv run escalane decide \
  --backend ollama \
  --model d3:4b \
  --require-live \
  --state src/escalane/examples/incidents/php_502.json \
  --catalog datasets/sysops_decide/catalog.yaml \
  --gate-config A3
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Trimmed output from that exact live command on the seed PHP 502 incident:&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fngghaew6383gg4ntirkq.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fngghaew6383gg4ntirkq.png" alt="escalane decide live Ollama d3:4b A3 terminal" width="800" height="782"&gt;&lt;/a&gt;&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight json"&gt;&lt;code&gt;&lt;span class="p"&gt;{&lt;/span&gt;
  &lt;span class="nl"&gt;"model"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="s2"&gt;"d3:4b"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
  &lt;span class="nl"&gt;"backend"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="s2"&gt;"ollama"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
  &lt;span class="nl"&gt;"playbook"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="s2"&gt;"restart_php_fpm"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
  &lt;span class="nl"&gt;"noul"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="mf"&gt;0.9878&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
  &lt;span class="nl"&gt;"blast_radius_score"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="mf"&gt;0.9822&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
  &lt;span class="nl"&gt;"blast_polarity"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="s2"&gt;"high_means_wide"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
  &lt;span class="nl"&gt;"gate_decision"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="s2"&gt;"hitl"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
  &lt;span class="nl"&gt;"gate_reason"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="s2"&gt;"label_precision"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
  &lt;span class="nl"&gt;"dry_run_only"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="kc"&gt;true&lt;/span&gt;
&lt;span class="p"&gt;}&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Read that as: live System One answers, catalog playbook selected, dry-run only, no shell args invented by the model, and A3 routed to human review instead of execute. &lt;strong&gt;&lt;code&gt;blast_radius_score&lt;/code&gt; high means wide / more disruptive&lt;/strong&gt;, not “safer” — &lt;code&gt;0.98&lt;/code&gt; is near the dangerous end of the ordinal. The CLI also emits &lt;code&gt;blast_polarity: "high_means_wide"&lt;/code&gt; so that legend cannot get lost.&lt;/p&gt;

&lt;p&gt;Public links:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;Repo: &lt;a href="https://github.com/escalane-lab/escalane" rel="noopener noreferrer"&gt;github.com/escalane-lab/escalane&lt;/a&gt;
&lt;/li&gt;
&lt;li&gt;Docs / site: &lt;a href="https://escalane-lab.github.io/escalane-docs/" rel="noopener noreferrer"&gt;escalane-lab.github.io/escalane-docs&lt;/a&gt;
&lt;/li&gt;
&lt;li&gt;Harness map: &lt;a href="https://github.com/escalane-lab/escalane-docs/blob/main/docs/harness-architecture.md" rel="noopener noreferrer"&gt;&lt;code&gt;docs/harness-architecture.md&lt;/code&gt;&lt;/a&gt;
&lt;/li&gt;
&lt;li&gt;System One wire format: &lt;a href="https://lmstudio.ai/docs/developer/jev-compat/systemone" rel="noopener noreferrer"&gt;Decisions (System One)&lt;/a&gt;
&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  Partner Technologies
&lt;/h2&gt;

&lt;p&gt;I did &lt;strong&gt;not&lt;/strong&gt; build on an MLH partner stack for this project — honesty over checkbox.&lt;/p&gt;

&lt;p&gt;What I actually used:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Python&lt;/strong&gt; + &lt;code&gt;uv&lt;/code&gt; for the &lt;code&gt;escalane&lt;/code&gt; CLI and harness&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;httpx&lt;/strong&gt; for HTTP to local decision backends&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Ollama / LM Studio&lt;/strong&gt; for local System One–style decision models (when the eGPU / APU cooperated)&lt;/li&gt;
&lt;li&gt;A &lt;strong&gt;custom Decide→Gates→Act harness&lt;/strong&gt; with a seed SysOps catalog and mock backend for offline plumbing&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;Challenges that felt real during the week: model pulls and DNS hiccups, an eGPU busy with chat models so decision models lived on a second box, and the constant trap of “CI green on mock” teaching wiring instead of decision quality. The claim fence and &lt;code&gt;--require-live&lt;/code&gt; exist so I stop lying to myself.&lt;/p&gt;

&lt;p&gt;We measured gate postures against ungated / free tool-calling baselines on a seed suite — enough to see that catalog membership alone is not enough under injection-shaped prompts. Full metrics stay out of this post; they belong in a later paper draft.&lt;/p&gt;

&lt;p&gt;This is a &lt;strong&gt;research harness&lt;/strong&gt;, not a production-certified remediator. Conservative gates will ask for a human more often than a demo agent that always “just runs the fix.”&lt;/p&gt;

&lt;h2&gt;
  
  
  Hackathon Experience
&lt;/h2&gt;

&lt;p&gt;Builders Week felt like permission to ship the &lt;em&gt;shape&lt;/em&gt; of the idea without pretending it was finished science. I built mostly solo, with community Discord as ambient company — other people grinding on their own weird stacks late into the night.&lt;/p&gt;

&lt;p&gt;What I’ll remember: sticky-note clarity. &lt;em&gt;The model chooses; the harness is the only thing that runs.&lt;/em&gt; Locking the CLI path (&lt;code&gt;decide&lt;/code&gt; / &lt;code&gt;run&lt;/code&gt; / &lt;code&gt;replay&lt;/code&gt;), a seed catalog, and the Decide→Gates→Act skeleton mattered more than chasing a flashy demo that shells out from a prompt.&lt;/p&gt;

&lt;p&gt;What matured after the week was still builder work: params-from-state &lt;strong&gt;plus&lt;/strong&gt; Act-time allowlists, gate postures that treat model confidence as a filter not a warrant, blast scores with an explicit polarity legend, and fail-closed live backends.&lt;/p&gt;

&lt;p&gt;If you are also fighting “LLM with bash privileges,” I would love to hear how you drew the trust boundary — especially where you put noun allowlists versus model self-report.&lt;/p&gt;

</description>
      <category>mlhacks</category>
      <category>ai</category>
      <category>python</category>
      <category>security</category>
    </item>
    <item>
      <title>Captain Hook: When Your AI Copilot Gets Promoted to Captain!</title>
      <dc:creator>Mátyás Kuti-Kreszács</dc:creator>
      <pubDate>Thu, 29 Jan 2026 22:00:54 +0000</pubDate>
      <link>https://dev.to/thec0dewriter/captain-hook-when-your-ai-copilot-gets-promoted-to-captain-3amc</link>
      <guid>https://dev.to/thec0dewriter/captain-hook-when-your-ai-copilot-gets-promoted-to-captain-3amc</guid>
      <description>&lt;p&gt;&lt;em&gt;This is a submission for the &lt;a href="https://dev.to/challenges/github-2026-01-21"&gt;GitHub Copilot CLI Challenge&lt;/a&gt;&lt;/em&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  What I Built
&lt;/h2&gt;

&lt;p&gt;Ahoy, matey! 🏴‍☠️ Let me tell you a tale about &lt;strong&gt;Captain Hook&lt;/strong&gt; – not the villain from Neverland, but the hero of your git repository!&lt;/p&gt;

&lt;p&gt;You see, GitHub Copilot usually sits beside you as a friendly &lt;strong&gt;copilot&lt;/strong&gt;, offering helpful suggestions. But what if your copilot got &lt;strong&gt;promoted to captain&lt;/strong&gt; and started giving you &lt;em&gt;orders&lt;/em&gt;? What if it didn't just suggest – it &lt;em&gt;enforced&lt;/em&gt;? &lt;/p&gt;

&lt;p&gt;That's exactly what I built: an intelligent git pre-commit &lt;strong&gt;hook&lt;/strong&gt; that acts like a &lt;strong&gt;captain&lt;/strong&gt;, commanding your code to ship-shape quality before it sets sail into your repository!&lt;/p&gt;

&lt;h3&gt;
  
  
  🎣 The Hook That Captains Your Code
&lt;/h3&gt;

&lt;p&gt;&lt;strong&gt;Captain Hook&lt;/strong&gt; is a git pre-commit hook with a personality problem (in the best way). It:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;📋 Linting&lt;/strong&gt; (ruff check) - &lt;em&gt;"Arr! Walk the plank, unused imports!"&lt;/em&gt;
&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;🎨 Formatting&lt;/strong&gt; (ruff format) - &lt;em&gt;"Straighten those lines, ye scurvy code!"&lt;/em&gt;
&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;🔎 Type Checking&lt;/strong&gt; (mypy) - &lt;em&gt;"Where be yer type hints, sailor?"&lt;/em&gt;
&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;🧪 Tests&lt;/strong&gt; (pytest) - &lt;em&gt;"No code leaves port without passing inspection!"&lt;/em&gt;
&lt;/li&gt;
&lt;/ul&gt;

&lt;h3&gt;
  
  
  ⚓ From Copilot to Captain
&lt;/h3&gt;

&lt;p&gt;Here's the twist that makes this project special:&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Normal Copilot&lt;/strong&gt;: "Hey, maybe you could fix this import?" 😊&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Captain Hook&lt;/strong&gt;: "HALT! Unused import detected! Removing it NOW!" 💪&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Normal Copilot&lt;/strong&gt;: "Consider adding type hints here..." 🤔&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Captain Hook&lt;/strong&gt;: "NO TYPE HINTS? Calling in backup! &lt;em&gt;blows whistle&lt;/em&gt; GitHub Copilot CLI, report for duty!" 🎺&lt;/p&gt;

&lt;p&gt;When Captain Hook encounters something it can't auto-fix (type errors, test failures), it doesn't just block your commit and walk away. Oh no! It calls in the &lt;strong&gt;GitHub Copilot CLI&lt;/strong&gt; – think of it as calling in the &lt;em&gt;Admiral&lt;/em&gt; for strategic advice!&lt;/p&gt;

&lt;h3&gt;
  
  
  🏴‍☠️ Why "Captain Hook"?
&lt;/h3&gt;

&lt;p&gt;Because this git &lt;strong&gt;hook&lt;/strong&gt; acts like a &lt;strong&gt;captain&lt;/strong&gt;:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Hook&lt;/strong&gt; (git pre-commit mechanism) → Technical term&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Captain&lt;/strong&gt; (commands and enforces) → How it behaves&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Captain Hook&lt;/strong&gt; → The perfect nautical pun! 🎣&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;Plus, just like the infamous Captain Hook, this tool:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;✅ Never lets Peter Pan (bad code) escape&lt;/li&gt;
&lt;li&gt;✅ Has a trusty sidekick (GitHub Copilot CLI)&lt;/li&gt;
&lt;li&gt;✅ Commands the crew (QA tools) with authority&lt;/li&gt;
&lt;li&gt;✅ Protects the ship (your repository) fiercely&lt;/li&gt;
&lt;/ul&gt;

&lt;h3&gt;
  
  
  🚢 What It Means to Me
&lt;/h3&gt;

&lt;p&gt;As developers, we often treat CI/CD and code quality tools as annoying gatekeepers – the grumpy pirates who say "NO!" to everything. But what if they were more like a wise captain who:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;Fixes what can be fixed automatically (80% of issues)&lt;/li&gt;
&lt;li&gt;Teaches you how to fix complex problems (using AI)&lt;/li&gt;
&lt;li&gt;Keeps the ship running smoothly (maintains quality)&lt;/li&gt;
&lt;li&gt;Actually makes your life &lt;em&gt;easier&lt;/em&gt;, not harder&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;&lt;strong&gt;Captain Hook&lt;/strong&gt; represents a shift from "code quality as punishment" to "code quality as a helpful mentor with a sense of humor and a pirate hat."&lt;/p&gt;

&lt;h2&gt;
  
  
  Demo
&lt;/h2&gt;

&lt;p&gt;🔗 &lt;strong&gt;Live Repository&lt;/strong&gt;: &lt;a href="https://github.com/thec0dewriter/captain_hook.git" rel="noopener noreferrer"&gt;https://github.com/thec0dewriter/captain_hook.git&lt;/a&gt;&lt;/p&gt;

&lt;h3&gt;
  
  
  🗺️ Project Repository Structure:
&lt;/h3&gt;



&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight plaintext"&gt;&lt;code&gt;captain_hook/
├── src/captain_hook/         # The treasure (Python library code)
├── tests/                    # Quality inspection (100% coverage)
├── hooks/                    # The Captain's quarters! ⚓
│   ├── pre-commit           # Captain Hook himself
│   ├── install.sh           # Recruit the captain
│   ├── demo.sh              # Training exercises
│   └── README.md            # Captain's orders
├── submission.md            # This very tale!
├── HOOK_SUMMARY.md          # The Captain's log
└── README.md                # Ship's manifest
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;h3&gt;
  
  
  ⚓ Captain Hook in Action:
&lt;/h3&gt;

&lt;p&gt;&lt;strong&gt;Example 1: The Captain Takes Command&lt;/strong&gt; 💪&lt;/p&gt;

&lt;p&gt;Picture this: You try to commit some messy code...&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;&lt;span class="nv"&gt;$ &lt;/span&gt;git commit &lt;span class="nt"&gt;-m&lt;/span&gt; &lt;span class="s2"&gt;"Add new feature"&lt;/span&gt;

🔍 Running pre-commit QA checks...

📋 Step 1/4: Linting &lt;span class="o"&gt;(&lt;/span&gt;ruff check&lt;span class="o"&gt;)&lt;/span&gt;
✗ Linting failed
  - Found unused imports: os, sys, json
  - Missing spaces around operators

⚓ &lt;span class="s2"&gt;"AVAST, YE SCURVY CODE!"&lt;/span&gt; - Captain Hook

Attempting auto-fix with ruff...
✓ Auto-fixed linting issues
✓ Unused imports walked the plank!
✓ Spaces restored to their rightful place!
Staging auto-fixed files...

🎨 Step 2/4: Format checking &lt;span class="o"&gt;(&lt;/span&gt;ruff format&lt;span class="o"&gt;)&lt;/span&gt;
✗ Format check failed

⚓ &lt;span class="s2"&gt;"This code be messier than a pirate's beard!"&lt;/span&gt;

Auto-formatting files...
✓ Files formatted
✓ Your code now looks shipshape!
Staging formatted files...

🔎 Step 3/4: Type checking &lt;span class="o"&gt;(&lt;/span&gt;mypy&lt;span class="o"&gt;)&lt;/span&gt;
✓ Type checking passed

⚓ &lt;span class="s2"&gt;"Good! Proper type hints, as any respectable sailor knows!"&lt;/span&gt;

🧪 Step 4/4: Running tests &lt;span class="o"&gt;(&lt;/span&gt;pytest&lt;span class="o"&gt;)&lt;/span&gt;
✓ Tests passed

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
✅ All pre-commit checks PASSED
⚓ &lt;span class="s2"&gt;"Permission granted! Your code may set sail!"&lt;/span&gt;
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;&lt;strong&gt;Example 2: When the Captain Calls for Backup&lt;/strong&gt; 🎺&lt;/p&gt;

&lt;p&gt;Sometimes even captains need help from the admiral...&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;&lt;span class="nv"&gt;$ &lt;/span&gt;git commit &lt;span class="nt"&gt;-m&lt;/span&gt; &lt;span class="s2"&gt;"Add calculator"&lt;/span&gt;

🔍 Running pre-commit QA checks...

📋 Step 1/4: Linting &lt;span class="o"&gt;(&lt;/span&gt;ruff check&lt;span class="o"&gt;)&lt;/span&gt;
✓ Linting passed

🎨 Step 2/4: Format checking &lt;span class="o"&gt;(&lt;/span&gt;ruff format&lt;span class="o"&gt;)&lt;/span&gt;
✓ Format check passed

🔎 Step 3/4: Type checking &lt;span class="o"&gt;(&lt;/span&gt;mypy&lt;span class="o"&gt;)&lt;/span&gt;
✗ Type checking failed
src/calculator.py:4: error: Function is missing a &lt;span class="nb"&gt;type &lt;/span&gt;annotation

⚓ &lt;span class="s2"&gt;"Blimey! Missing type hints be beyond me fixing powers!"&lt;/span&gt;
⚓ &lt;span class="s2"&gt;"Calling in the ADMIRAL for strategic advice!"&lt;/span&gt;

⚠️ Attempting auto-fix with GitHub Copilot CLI...

🎺 &lt;span class="k"&gt;*&lt;/span&gt;Admiral GitHub Copilot CLI reports &lt;span class="k"&gt;for &lt;/span&gt;duty&lt;span class="k"&gt;*&lt;/span&gt;

💡 Copilot suggests:
   &lt;span class="s2"&gt;"Add type annotations to function parameters and return type:

    def calculate(a: float, b: float) -&amp;gt; float:
        return a + b"&lt;/span&gt;

⚓ &lt;span class="s2"&gt;"There ye have it! Follow the Admiral's orders, sailor!"&lt;/span&gt;

Tips:
  - Apply the Admiral&lt;span class="s1"&gt;'s wisdom (Copilot'&lt;/span&gt;s suggestion&lt;span class="o"&gt;)&lt;/span&gt;
  - Run &lt;span class="s1"&gt;'make check'&lt;/span&gt; to verify
  - Try committing again
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;h3&gt;
  
  
  🏴‍☠️ Real Voyage Testing
&lt;/h3&gt;

&lt;p&gt;The Captain was battle-tested with intentionally broken code:&lt;/p&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Challenge&lt;/th&gt;
&lt;th&gt;Captain's Response&lt;/th&gt;
&lt;th&gt;Outcome&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;Unused imports (3x)&lt;/td&gt;
&lt;td&gt;"Walk the plank!"&lt;/td&gt;
&lt;td&gt;✅ Auto-removed&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Bad spacing &lt;code&gt;x,y&lt;/code&gt;
&lt;/td&gt;
&lt;td&gt;"Mind yer spaces!"&lt;/td&gt;
&lt;td&gt;✅ Fixed to &lt;code&gt;x, y&lt;/code&gt;
&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Messy formatting&lt;/td&gt;
&lt;td&gt;"Swab the deck!"&lt;/td&gt;
&lt;td&gt;✅ Auto-formatted&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Missing type hints&lt;/td&gt;
&lt;td&gt;"Call the Admiral!"&lt;/td&gt;
&lt;td&gt;✅ Copilot suggested fix&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Code ready to commit&lt;/td&gt;
&lt;td&gt;"Set sail!"&lt;/td&gt;
&lt;td&gt;✅ Smooth sailing&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;h3&gt;
  
  
  🗺️ Quick Start: Recruit Your Captain
&lt;/h3&gt;



&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;&lt;span class="c"&gt;# Clone the ship&lt;/span&gt;
git clone https://github.com/thec0dewriter/captain_hook.git
&lt;span class="nb"&gt;cd &lt;/span&gt;captain_hook

&lt;span class="c"&gt;# Stock the supplies (install dependencies)&lt;/span&gt;
make &lt;span class="nb"&gt;install&lt;/span&gt;

&lt;span class="c"&gt;# Recruit Captain Hook to your crew&lt;/span&gt;
make install-hooks

&lt;span class="c"&gt;# Make a change and try to commit&lt;/span&gt;
&lt;span class="nb"&gt;echo&lt;/span&gt; &lt;span class="s2"&gt;"import os"&lt;/span&gt; &lt;span class="o"&gt;&amp;gt;&amp;gt;&lt;/span&gt; src/captain_hook/calculator.py
git add &lt;span class="nb"&gt;.&lt;/span&gt;
git commit &lt;span class="nt"&gt;-m&lt;/span&gt; &lt;span class="s2"&gt;"Test the Captain"&lt;/span&gt;

&lt;span class="c"&gt;# Watch Captain Hook spring into action! 🎣&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Farticles%2Fiqshb4v0ai3n50l7mko6.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Farticles%2Fiqshb4v0ai3n50l7mko6.png" alt="Working screenshot" width="800" height="836"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  My Experience with GitHub Copilot CLI
&lt;/h2&gt;

&lt;h3&gt;
  
  
  🎭 The Plot Twist: From Copilot to Captain
&lt;/h3&gt;

&lt;p&gt;When I started this project, I thought I was building a simple pre-commit hook. But something magical happened: &lt;strong&gt;the copilot became the captain&lt;/strong&gt;.&lt;/p&gt;

&lt;p&gt;Here's how GitHub Copilot CLI transformed this project:&lt;/p&gt;

&lt;h3&gt;
  
  
  1️⃣ &lt;strong&gt;The Captain's Wisdom: Intelligent Error Resolution&lt;/strong&gt;
&lt;/h3&gt;

&lt;p&gt;&lt;strong&gt;Before Copilot CLI:&lt;/strong&gt;&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight plaintext"&gt;&lt;code&gt;Error: Function missing type annotation
*stares at screen*
*Googles "python type hints"*
*reads docs for 10 minutes*
*tries something*
*fails*
*repeats*
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;&lt;strong&gt;With Captain Hook + Copilot CLI:&lt;/strong&gt;&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight plaintext"&gt;&lt;code&gt;⚓ "Type hints be missing! Calling the Admiral!"
💡 Copilot: "Add type hints like this: def foo(x: int) -&amp;gt; int:"
✅ "Brilliant! Fixed in 30 seconds!"
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;It's like having a &lt;strong&gt;ship's navigator&lt;/strong&gt; who actually knows where you're going instead of just pointing at the horizon!&lt;/p&gt;

&lt;h3&gt;
  
  
  2️⃣ &lt;strong&gt;Building the Captain: Meta-Development&lt;/strong&gt;
&lt;/h3&gt;

&lt;p&gt;The beautiful irony? I used GitHub Copilot CLI to build a tool that &lt;em&gt;uses&lt;/em&gt; GitHub Copilot CLI!&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Inception moment:&lt;/strong&gt;&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;Used Copilot to write the hook&lt;/li&gt;
&lt;li&gt;The hook calls Copilot CLI for help&lt;/li&gt;
&lt;li&gt;It's copilots all the way down! 🎭&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;Copilot helped me:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;Scaffold the entire git hook structure&lt;/li&gt;
&lt;li&gt;Write robust bash error handling&lt;/li&gt;
&lt;li&gt;Design the captain personality (yes, the pirate puns!)&lt;/li&gt;
&lt;li&gt;Generate comprehensive documentation&lt;/li&gt;
&lt;li&gt;Create engaging demo scenarios&lt;/li&gt;
&lt;/ul&gt;

&lt;h3&gt;
  
  
  3️⃣ &lt;strong&gt;The Admiral's Strategy: Learning Through Collaboration&lt;/strong&gt;
&lt;/h3&gt;

&lt;p&gt;Working with Copilot CLI felt less like "using a tool" and more like &lt;strong&gt;pair programming with a pirate crew&lt;/strong&gt;:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;The Captain&lt;/strong&gt; (my hook) → Commands and enforces&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;The Admiral&lt;/strong&gt; (Copilot CLI) → Provides strategic advice&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;The Crew&lt;/strong&gt; (ruff, mypy, pytest) → Do the actual work&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Me&lt;/strong&gt; (the developer) → Enjoys smooth sailing! ⛵&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;Best part? When Captain Hook calls Copilot CLI, it's not just showing error messages – it's providing &lt;strong&gt;contextual wisdom&lt;/strong&gt; based on the actual code and error.&lt;/p&gt;

&lt;h3&gt;
  
  
  4️⃣ &lt;strong&gt;Real-World Impact: The 80/20 Rule of Joy&lt;/strong&gt;
&lt;/h3&gt;

&lt;p&gt;Here's the magic formula Captain Hook delivers:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;80% of problems&lt;/strong&gt;: Auto-fixed instantly (bye bye, formatting fights!)&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;20% of problems&lt;/strong&gt;: Copilot CLI teaches you the solution&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;100% of the time&lt;/strong&gt;: You stay in flow state 🌊&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;&lt;strong&gt;Old workflow:&lt;/strong&gt;&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight plaintext"&gt;&lt;code&gt;Write code → Commit → CI fails → Context switch → Google → Fix → Repeat
😤 Frustration level: High
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;&lt;strong&gt;Captain Hook workflow:&lt;/strong&gt;&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight plaintext"&gt;&lt;code&gt;Write code → Commit attempt → Captain fixes most issues → 
Copilot teaches remaining fixes → Commit succeeds
😊 Frustration level: "What frustration?"
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;h3&gt;
  
  
  5️⃣ &lt;strong&gt;The Joke That Became Real&lt;/strong&gt;
&lt;/h3&gt;

&lt;p&gt;Started as: "Haha, what if git hooks had personality?"&lt;/p&gt;

&lt;p&gt;Became: "Wait, this is actually the future of developer experience!"&lt;/p&gt;

&lt;p&gt;The &lt;strong&gt;Captain Hook&lt;/strong&gt; name started as a pun but revealed a deeper truth:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;Copilots &lt;em&gt;suggest&lt;/em&gt; (helpful but passive)&lt;/li&gt;
&lt;li&gt;Captains &lt;em&gt;command&lt;/em&gt; (authoritative but caring)&lt;/li&gt;
&lt;li&gt;The best tools do both! 🎯&lt;/li&gt;
&lt;/ul&gt;

&lt;h3&gt;
  
  
  🎬 Key Takeaway: It's About the Journey
&lt;/h3&gt;

&lt;p&gt;GitHub Copilot CLI isn't just "AI that writes code faster." It's:&lt;/p&gt;

&lt;p&gt;✨ &lt;strong&gt;A teaching tool&lt;/strong&gt; that explains &lt;em&gt;why&lt;/em&gt; fixes work&lt;br&gt;
🚢 &lt;strong&gt;A workflow enhancer&lt;/strong&gt; that keeps you sailing smoothly&lt;br&gt;&lt;br&gt;
🎣 &lt;strong&gt;A captain&lt;/strong&gt; that catches problems before they cause shipwrecks&lt;br&gt;
💡 &lt;strong&gt;A collaboration partner&lt;/strong&gt; that makes coding more fun&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Captain Hook proves that AI in development isn't about replacing developers – it's about promoting your helpful copilot to a wise captain who still lets you steer the ship!&lt;/strong&gt;&lt;/p&gt;


&lt;h3&gt;
  
  
  🎯 Technologies Used:
&lt;/h3&gt;

&lt;ul&gt;
&lt;li&gt;Python 3.9+ with UV package manager (fastest ship in the fleet!)&lt;/li&gt;
&lt;li&gt;Ruff (linting &amp;amp; formatting) (the ship's cleaning crew)&lt;/li&gt;
&lt;li&gt;Mypy (type checking) (the quality inspector)&lt;/li&gt;
&lt;li&gt;Pytest (testing with coverage) (safety drills)&lt;/li&gt;
&lt;li&gt;Bash scripting (the ship's machinery)&lt;/li&gt;
&lt;li&gt;Git hooks (the captain's perch)&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;GitHub Copilot CLI&lt;/strong&gt; (the wise admiral) 🎖️&lt;/li&gt;
&lt;/ul&gt;
&lt;h3&gt;
  
  
  📋 Current Status &amp;amp; Future Voyages
&lt;/h3&gt;

&lt;p&gt;&lt;strong&gt;✅ What's Shipshape:&lt;/strong&gt;&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;Intelligent pre-commit hook with auto-fix&lt;/li&gt;
&lt;li&gt;GitHub Copilot CLI integration&lt;/li&gt;
&lt;li&gt;100% test coverage (8 passing tests)&lt;/li&gt;
&lt;li&gt;Comprehensive documentation&lt;/li&gt;
&lt;li&gt;Real-world testing completed&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;&lt;strong&gt;🚧 Potential Future Adventures (TODOs):&lt;/strong&gt;&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;[ ] Add more "captain personality" responses for different error types&lt;/li&gt;
&lt;li&gt;[ ] Support for commit-msg hook (Captain checks your commit messages!)&lt;/li&gt;
&lt;li&gt;[ ] Integration with more language ecosystems (TypeScript, Go, Rust)&lt;/li&gt;
&lt;li&gt;[ ] "Parrot mode" - hook verbosity level (quiet parrot vs. chatty parrot 🦜)&lt;/li&gt;
&lt;li&gt;[ ] Collectible "achievements" when you fix issues&lt;/li&gt;
&lt;li&gt;[ ] Ship's log (detailed commit history with captain's commentary)&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;&lt;strong&gt;Try it yourself:&lt;/strong&gt;&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;git clone https://github.com/thec0dewriter/captain_hook.git
&lt;span class="nb"&gt;cd &lt;/span&gt;captain_hook
make &lt;span class="nb"&gt;install
&lt;/span&gt;make install-hooks

&lt;span class="c"&gt;# Let Captain Hook command your code to quality! ⚓&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;






&lt;p&gt;&lt;strong&gt;Built with ❤️ (and lots of pirate puns) for the GitHub Copilot CLI Challenge&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;&lt;em&gt;"Not all those who wander are lost, but all code that wanders needs a captain!"&lt;/em&gt; 🎣⚓&lt;/p&gt;

</description>
      <category>devchallenge</category>
      <category>githubchallenge</category>
      <category>cli</category>
      <category>githubcopilot</category>
    </item>
  </channel>
</rss>
