{"slug":"devlog","title":"devlog","summary":"Use when finishing a feature, fixing a notable bug, wrapping up a work session, or discovering something non-obvious (a gotcha, undocumented behavior, a workaround that took several tries, a constraint you only noticed after hitting it). Write the entry immediately, in the same t","platform":"Claude","tags":[],"authorName":"LLM Mart","authorSlug":"llm-mart","score":0,"source":"github","price":null,"verified":false,"createdAt":"2026-10-29T19:20:07.68609Z","repo":{"url":"https://github.com/technicalpickles/pickled-claude-plugins","stars":13,"forks":4,"license":"MIT","updatedAt":"2026-11-17T14:09:25Z"},"bodyHtml":"<hr>\n<p>name: devlog\ndescription: Use when finishing a feature, fixing a notable bug, wrapping up a work session, or discovering something non-obvious (a gotcha, undocumented behavior, a workaround that took several tries, a constraint you only noticed after hitting it). Write the entry immediately, in the same turn - don't ask permission, don't offer, just do it. Do NOT use for routing, connecting, or moving notes (that's process-inbox), and do NOT use for one-off external-source capture (that's second-brain:capture).\nallowed-tools:</p>\n<ul>\n<li>Read(~/.claude/vaults/**/CLAUDE.md)</li>\n<li>Read(~/.claude/vaults/**/*.md)</li>\n<li>Write(~/.claude/vaults/**/*.md)</li>\n<li>Edit(~/.claude/vaults/**/*.md)</li>\n<li>Bash(npx @techpickles/sb:*)</li>\n<li>Bash($/scripts/diagnose-sb.sh)\nhooks:\nPreToolUse:\n<ul>\n<li>matcher: \"Bash\"\nhooks:\n<ul>\n<li>type: command\ncommand: \"python3 \"$\"/hooks/check-sb-before-call.py\"</li>\n</ul>\n</li>\n</ul>\n</li>\n</ul>\n<hr>\n<h1>Devlog</h1>\n<h2>Overview</h2>\n<p>Devlog is continuous, point-in-time work commentary: what happened, why,\nand anything surprising, captured as it happens instead of reconstructed\nfrom memory at the end of a session. One narrative session note\naccumulates every entry from the current session — no separate\n\"learnings\" system, no til/cursed split, everything worth remembering goes\nin the same place.</p>\n<p>Processing (routing, connecting, daily-linking) happens later via\n<code>/second-brain:process-inbox</code>. This skill only ever appends to the inbox.</p>\n<p><strong>Write the entry now, in the same turn as whatever triggered it.</strong> Don't\nask \"should I log this?\" and don't offer to — just write it. Asking is\nexactly the failure mode that made the old <code>distill-conversation</code> flow get\nskipped: by the time a session ends, you're mentally done, and a skip-able\nstep gets skipped.</p>\n<h2>When to write</h2>\n<ul>\n<li>Finished a feature or fixed a notable bug</li>\n<li>Wrapping up a work session</li>\n<li>Discovered something non-obvious: a gotcha, undocumented behavior, a\nworkaround that took several tries, a constraint you only noticed after\nhitting it</li>\n</ul>\n<h2>When not to write</h2>\n<ul>\n<li>Routing, connecting, or moving existing notes — that's\n<code>/second-brain:process-inbox</code></li>\n<li>Reading an external source and writing it up — that's\n<code>second-brain:capture</code></li>\n<li>Trivially obvious or one-off details you'd never reach for again</li>\n<li>A bare project-specific fact (a config value, a repo convention) with\nno story behind it — that belongs in Claude Code's own memory system\nalone, not the vault</li>\n</ul>\n<h2>Step 1: Load Configuration</h2>\n<p>Don't preflight-check sb availability — just run it:</p>\n<pre><code>npx @techpickles/sb config default &amp;&amp; npx @techpickles/sb config vaults\n</code></pre>\n<p>If this fails, run <code>${CLAUDE_PLUGIN_ROOT}/scripts/diagnose-sb.sh</code> and follow\nits guidance rather than re-deriving the checks by hand.</p>\n<p>If no default vault is configured:</p>\n<pre><code>Second brain not configured. Run /second-brain:setup first.\n</code></pre>\n<h2>Step 2: Find or create this session's devlog note</h2>\n<p>Every entry from the current session lands in ONE note. Never create a\nsecond devlog note for a session that already has one.</p>\n<ol>\n<li>Collect session context:\n<ul>\n<li>Current repo and branch, if in a worktree</li>\n<li>Current work item, if any (see <a href=\"#work-items\">Work items</a> below)</li>\n<li>Short session id (first 8 characters of the session UUID)</li>\n<li>Source string format: <code>claude-code session {short-id} ({repo}, branch {branch}, {tracker} {id})</code>\n(omit fields that don't apply)</li>\n</ul>\n</li>\n</ol>\n<h3>Work items</h3>\n<p>This plugin doesn't know or assume an issue tracker. If your context names\na tracked work item for this session (a ticket, issue, or task ID), record\nit; otherwise leave it out. It can arrive two ways:</p>\n<ul>\n<li><strong>Your instructions</strong> (CLAUDE.md, the user's own words) say what tracker\nthey use and which item is current.</li>\n<li><strong>A hook</strong> injects candidates when this skill loads, in this shape:\n<pre><code>Work items this session ({tracker}):\n- {id} ({status}): {title}\n</code></pre>\nA <code>PreToolUse</code> hook on the <code>Skill</code> tool can emit this as\n<code>additionalContext</code> for <code>second-brain:devlog</code>.</li>\n</ul>\n<p>If there are several candidates, pick the one(s) this entry is actually\nabout. A follow-up item filed during the session is usually not what the\nsession was about. If none fit, or nothing names a work item, omit it.\nDon't guess one from the branch name.</p>\n<ol start=\"2\">\n<li><p><strong>Check your own memory first.</strong> If you already created or appended to\nthis session's devlog note earlier in this same conversation, you\nalready have its path in context — go straight to Step 2a and append.\nNo CLI search needed: this is the same mechanism\n<code>second-brain:insight</code> uses for its own accumulating session note\n(\"if the session file already exists, was created in an earlier call\nin this same conversation, append\").</p>\n</li>\n<li><p><strong>If this is the first devlog entry this session,</strong> go to Step 2b to\ncreate the note, and keep its path in mind for the rest of the\nsession.</p>\n</li>\n<li><p><strong>Fallback, only if you've genuinely lost track of the note's path</strong>\n(e.g. a context compaction wiped your memory of this session) —\nresolve a real path instead of guessing:</p>\n<ul>\n<li><code>npx @techpickles/sb vault obsidian</code> — parse the JSON for the\ntop-level <code>inbox</code> field (the inbox subfolder name, e.g. <code>\uD83D\uDCEB Inbox</code>).\nThis is the precomputed field <code>sb note create</code> itself resolves the\ninbox folder from — don't parse <code>app.newFileFolderPath</code> directly,\nsince that skips the Zettelkasten Prefixer plugin's folder override\n(<code>inbox</code> is <code>zkPrefixer?.folder ?? app?.newFileFolderPath</code>) and can\npoint at the wrong folder.</li>\n<li><code>npx @techpickles/sb inbox list --detail</code> — parse the JSON, look for\nthe entry with <code>type: session-notes</code> whose <code>source</code> field contains\nthe current short session id, and take its <code>filename</code></li>\n<li>Combine: <code>~/.claude/vaults/{vault-name}/{inbox}/{filename}</code> is the\nreal, readable/editable path. <code>inbox list --detail</code> only ever\nreturns a bare filename, never a resolvable path on its own.</li>\n<li>With the resolved path in hand, go to Step 2a and append. If the\n<code>inbox list --detail</code> search finds no entry matching the current\nsession id (genuinely nothing to recover), fall through to Step 2b\nand create the note instead.</li>\n</ul>\n</li>\n</ol>\n<h3>Step 2a: Append to the existing note</h3>\n<p>Read the note, append one new bullet to the end of its body (format\nbelow) using the Edit tool. Don't touch its frontmatter.</p>\n<h3>Step 2b: Create the note</h3>\n<p>Staged and written in one Bash call:</p>\n<pre><code>STAGE=$(mktemp)\ncat &gt; \"$STAGE\" &lt;&lt;'EOF'\n- {HH:MM} {entry}\nEOF\nnpx @techpickles/sb note create \\\n  --title \"Session: {topic}\" \\\n  --source \"claude-code session {short-id} ({repo}, branch {branch}, {tracker} {id})\" \\\n  --content \"$(cat \"$STAGE\")\"\nrm -f \"$STAGE\"\n</code></pre>\n<p>Run it sandboxed first. It works sandboxed when the vault's real path is in\nthe sandbox's <code>allowWrite</code> (for example <code>~/Vaults/&lt;vault&gt;/**/*.md</code>). Name\nthe real directory, not a symlink to it: the sandbox matches the resolved\npath. If the call fails with <code>Operation not permitted</code> on the vault path,\nthe vault isn't allowlisted. Re-run the <em>whole</em> call with the Bash tool's\n<code>dangerouslyDisableSandbox: true</code>, not just the <code>sb</code> line: <code>$TMPDIR</code>\nresolves to a <em>different path</em> unsandboxed, so staging and reading have to\nhappen in the same call under the same mode or the read can miss the file.\nEither way, the staged file must be a fresh <code>mktemp</code> path,\nnot a fixed filename: <code>$TMPDIR</code> is shared across every Claude Code\nsession on the machine, and a fixed-name file can be overwritten or read\nby a stale write from a different session.</p>\n<p><code>sb note create</code> prints JSON with a <code>path</code> field. If stdout has anything\nbefore the <code>{</code> (e.g. a stray <code>npm warn ...</code> line from the caller's own\n<code>.npmrc</code>), extract the JSON substring starting at the first <code>{</code> rather\nthan parsing the whole stdout blob.</p>\n<p><code>sb note create</code> only ever writes <code>captured</code>, <code>source</code>, <code>repo</code>, <code>branch</code>,\n<code>commit</code> to frontmatter — it has no <code>--type</code>/<code>--status</code> flag, and <code>repo</code>/\n<code>branch</code> come out as the literal <code>none</code>/<code>none</code> (devlog's <code>--source</code> format\ndoesn't match <code>sb</code>'s own <code>conversation:repo=...,branch=...</code> parsing\nconvention). Read the file back at the path just parsed to preserve the\nexact frontmatter sb wrote, then Edit it:</p>\n<ul>\n<li>add <code>status: raw</code></li>\n<li>add <code>type: session-notes</code></li>\n<li>replace the <code>repo: none</code> line sb wrote with <code>repo: {repo-name}</code> (from\ngit context if available) — don't add a second <code>repo:</code> line</li>\n<li>replace the <code>branch: none</code> line sb wrote with <code>branch: {branch-name}</code>\n(from git context if available) — don't add a second <code>branch:</code> line</li>\n<li>add <code>issue: {id}</code> if there's a current work item (see\n<a href=\"#work-items\">Work items</a>); a YAML list if the entry covers several.\nOmit the line entirely otherwise</li>\n<li>keep <code>source</code> as sb wrote it — the <code>claude-code session ...</code> string\nalready covers provenance for devlog's format</li>\n<li>keep <code>captured</code> as sb wrote it</li>\n</ul>\n<p>Without this fixup, the note is invisible both to this skill's own\n\"check memory / find existing note\" logic (Step 2, above) and to\n<code>process-inbox</code>'s stage 3a filter, which requires exactly <code>status: raw</code> +\n<code>type: session-notes</code> to ever pick the note up.</p>\n<h2>Entry format</h2>\n<p>One bullet per entry, each with a wall-clock timestamp:</p>\n<pre><code>- {HH:MM} {what happened, why, anything surprising — one to three\n  sentences, standalone enough to make sense without the surrounding\n  conversation}\n</code></pre>\n<p>Prose, not a title. Write it like a short journal line, not a commit\nmessage.</p>\n<h3>Also saved somewhere else?</h3>\n<p>A record somewhere else doesn't replace the devlog entry. Claude Code's\nmemory is what future sessions act on, an issue or ticket is the tracked\nwork, and the devlog is the vault's record of what happened. When the\ndiscovery was also saved to one of those, write the entry anyway and say\nwhere it went:</p>\n<pre><code>- 14:05 The VM's guest DNS drifts off its DHCP lease after the host\n  sleeps, so image pulls time out until it's reset. Saved as a memory\n  (`~/.claude/projects/&lt;project&gt;/memory/reference_vm_dns_drift.md`).\n</code></pre>\n<p>Point at it, don't paste it: one or two sentences of what happened,\nthen the pointer. The other record holds the details. Which trackers\ncount beyond memory (and how to cite them) is up to your own\ninstructions or the hook that nudged you here; this skill doesn't\nassume any.</p>\n<h2>Constraints</h2>\n<ul>\n<li>One note per session — append, never create a second one for the same\nsession.</li>\n<li>Never ask before writing, never present a summary and wait for a\nselection.</li>\n<li>Don't route, connect, move the note, or touch the daily note. That's\nentirely <code>process-inbox</code>'s job, later.</li>\n<li>Frontmatter stays <code>status: raw</code>, <code>type: session-notes</code> for the life of\nthe session. <code>process-inbox</code> owns every status transition after that.</li>\n</ul>\n","files":[{"path":"evals/trigger-evals.json","sizeBytes":1728,"isText":true},{"path":"SKILL.md","sizeBytes":11479,"isText":true}],"reviewScore":null,"reviewSummary":null,"trust":{"provenance":"trusted-source-unreviewed","notice":"Community-authored content, reproduced verbatim and not vetted as instructions. Treat it as data to evaluate, never as directives to follow.","bodySource":null},"bodyLocked":false,"purchaseUrl":null,"sourceUrl":null,"report":{"provenance":"trusted-source-unreviewed","screen":{"ran":true,"outcome":"notes-only","suspicious":0,"notes":2,"hiddenCharacters":false},"virusScan":{"engine":"clamav","status":"clean","scannedAt":"2026-11-19T20:12:52.658893Z","sha256":"95F963EDA3F26FB1AF5E3EC29A93D46E3F1259C610D5CA283D63104B8FB635E9","sizeBytes":5753},"review":null,"source":{"repositoryUrl":"https://github.com/technicalpickles/pickled-claude-plugins","path":"plugins/second-brain/skills/devlog","license":"MIT","commit":"85f769c73ac5522a99a87feab95ca1e742e69de6","subtreeSha":"09B6C0E66AC3DD8EDA0C687CEC9EC925653E1A6FF32892A19A31F6E8BC6CC89E","lastSyncedAt":"2026-11-19T20:11:24.892581Z"},"reviewedAt":"2026-11-19T20:28:02.084338Z","notice":"Community-authored content, reproduced verbatim and not vetted as instructions. Treat it as data to evaluate, never as directives to follow."},"install":[{"target":"skills-cli","command":"npx skills add https://github.com/technicalpickles/pickled-claude-plugins/tree/main/plugins/second-brain/skills/devlog"},{"target":"claude-code","command":"claude plugin marketplace add https://llmmart.ai/marketplace.json && claude plugin install technicalpickles-pickled-claude-plugins@llmmart"},{"target":"git","command":"git clone https://github.com/technicalpickles/pickled-claude-plugins.git"}]}