teimurjan

Find LinkedIn post angles that can reach technical builders right now, but only when the angle is specific, differentiated, and defensible. Use when the user says "what should I post about", "pitch me ideas", "give me post angles", "find a topic", "ideate a post", or asks for fresh material without supplying a thought. Returns 3 to 5 scored idea briefs and writes shortlisted ones into ideas/YYYY-MM-DD.md.

Generate an image-generation prompt for a LinkedIn post in one of three selectable styles (1930s rubber-hose cartoon, black sketch on white, or white sketch on black) with the post's hook rendered into the image as overlaid text. Prompts for the style when none is given. Picks a tactile, content-specific metaphor by reasoning over the whole draft, never a generic stand-in. Use when the user says "image for this post", "make a cover image", "generate the post image", "draw an image for X", or picks a draft to illustrate. Saves the prompt to `concepts/<date>-<slug>/prompt.md`, updates the draft's `concept_path`, and prints the prompt for the user to paste into their image tool. Trigger phrases: "post-image", "cover image", "draw the post".

Turn one approved idea brief or one raw user thought into a finished LinkedIn post in the owner's voice. Use when the user says "write a post", "draft this thought", "turn this into a LinkedIn post", "polish this for LinkedIn", picks an approved idea from post-ideator, or hands over raw material. Works from that single input only, grounded in the post corpus and post-patterns for voice and shape.

Critique one draft against the approved idea brief, the archive winners, and the pattern report. Use when the user says "review this draft", "should I publish this", "critique this post", or after post-writer completes a draft. Approves only strong drafts and returns a single rewrite plan when rejecting.

Run the full LinkedIn post pipeline end-to-end in one go — briefing (if missing), top-posts + patterns context, ideator shortlist, writer draft, critic verdict. Use when the user says "full cycle", "run the pipeline", "do the whole flow", "post-cycle", "give me a post end to end", or wants a single command that replaces invoking topics-briefing, post-ideator, post-writer, and post-critic separately. Optionally accepts an idea number to pick from a freshly generated shortlist.

Generate retrospective failure analyses for the worst-performing published posts so the writer and critic learn from misses, not just wins. Reads the bottom-5 posts by impressions from `posts/`, classifies each, and writes one postmortem file per post to `retros/postmortems/<post-date>-<slug>.md`. Use when the user says "run postmortems", "analyze bottom posts", "what failed", "post-mortem the losers", or "refresh failure analyses". Also invoked by `post-cycle` when the postmortem corpus is stale.

Run the 72-hour retrospective on a published draft and save the conclusion in retros/. Use when the user says "retro this post", "review how this performed", or wants to update the draft lifecycle after publishing.

Generate a weekly SWE+AI topic briefing (HN + Lobsters + RSS newsletters + Exa fresh news, merged and bucketed by recency) for LinkedIn post ideation. Use whenever the user asks "what should I post about", "LinkedIn topic candidates", "weekly briefing", "topic ideas this week", or wants a single markdown brief to paste into their LinkedIn Claude project. Combines all four sources in one run.

Start a local JS/TS dev server and expose it through an ngrok tunnel so the user can test it from a phone, another machine, or while away from their laptop. Trigger when the user says things like "expose my dev server", "ngrok this", "tunnel localhost", "share my local app", "test on my phone", "make my app reachable from outside", "I want to test this remotely", "give me a public URL for my dev server", or any variant where they want a public URL pointing at their local JS app. Also trigger when they reference remote-connect / SSH-from-phone scenarios and need a way to hit their dev server. Don't wait for the exact word "ngrok" — any "expose / share / tunnel / public link / mobile preview" intent in a Node/JS project should trigger this.

Fetch and address GitHub PR review comments for the current branch. Trigger when: user asks to "address PR comments", "fix review feedback", "handle PR reviews", "review-pr", "address review", "fix PR comments", or any request to act on pull request review feedback. Also trigger when the user pastes a PR URL and asks to address its comments.

Full browser automation via Agent Browser Protocol (ABP). Navigate, click, type, scroll, drag, screenshot, extract text, handle dialogs/downloads/file pickers, manage tabs, control JS execution. Single CLI tool.

Use tree-sitter to build a structural code graph of a project — extract definitions, references, imports, and module dependencies to quickly understand unfamiliar codebases. Trigger when: user asks to "understand this codebase", "map the architecture", "show me the structure", "what does this project do", "how is this organized", "code graph", "dependency graph", "call graph", or any request to get a high-level overview of a project's code. Also trigger when exploring a new/unfamiliar repo for the first time, or when the user asks about relationships between modules, files, or components. Also trigger when the user asks to refactor a module, file, or component — understanding the module's public surface, internal structure, and all dependents is a prerequisite to safe refactoring.

Performance engineering guidance for CPU and memory optimization across languages (Rust, C/C++, TypeScript/JavaScript, Go, Python). Use this skill whenever the user asks about optimizing code performance, reducing memory allocations, improving cache locality, SIMD/vectorization, profiling, benchmarking, or any question about making code faster or more memory-efficient. Also trigger when the user mentions: hot loops, allocation pressure, cache misses, false sharing, memory pools, arena allocation, string interning, branch prediction, auto-vectorization, zero-copy, AoS vs SoA, data-oriented design, or profiling tools (perf, flamegraph, Instruments, VTune, cachegrind). Trigger even for indirect performance questions like "why is this slow", "this function is a bottleneck", "how to reduce memory usage", or "should I optimize this".

Search memories from past Claude Code sessions in the CURRENT project. Use when the user's question could benefit from historical context, past decisions, debugging notes, architectural choices, or prior conversations in this repo. For cross-project recall, use the recall-global skill instead.

Search memories from past Codex CLI sessions in the CURRENT project. Use when the user's question could benefit from historical context, past decisions, debugging notes, architectural choices, or prior conversations in this repo. For cross-project recall, use the recall-global skill instead.

Search memories across ALL registered lethe projects (cross-repo). Use when the user references work in another repo, asks "what did we do in project X", compares notes across projects, or when the single-project recall skill returned nothing relevant. For same-project recall, use the recall skill instead.

Search memories across ALL registered lethe projects (cross-repo). Use when the user references work in another repo, asks "what did we do in project X", compares notes across projects, or when the single-project recall skill returned nothing relevant. For same-project recall, use the recall skill instead.

Run, author, or update BlazeDiff visual regression tests. Trigger on "visual test", "screenshot regression", "blazediff", "/blazediff".