[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M77KCWHZGRKPTN1KGD0WZ9VG":3,"skill-same:01M77KCWHZGRKPTN1KGD0WZ9VG":48},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":18,"publishedAt":19,"linkOnly":20,"markdownContent":21,"markdownBody":22,"frontMatter":23,"refName":24,"commitSha":25,"skillDirectory":26,"forksCount":27,"repositoryHtmlUrl":28,"repositoryHomepageUrl":29,"ownerHtmlUrl":30,"ownerBlogUrl":31,"ownerDescription":32,"updatedAt":33,"install":34,"resources":47},"01M77KCWHZGRKPTN1KGD0WZ9VG","optimize-skills",null,"Compile your local stellar-build usage traces into sharper skills. Use when the user says \"optimize my skills\", \"optimize-skills\", \"improve my skills from usage\", \"learn from my traces\", \"stellar-loop optimize\", or wants the learning loop to refine installed skills based on how they've actually been used. DSPy-style prompt optimization over the local trace store at ~\u002F.stellar\u002Ftraces - fully local, reversible.","WritzProtocol\u002Fwritz","WritzProtocol","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F296057967?v=4","01M77K5N0T3073T814APZTAXM4","Writz Protocol","01M77K5N1WQ551ZKCCK9MR379A",6,"Apache-2.0","https:\u002F\u002Fgithub.com\u002FWritzProtocol\u002Fwritz\u002Fblob\u002Fmain\u002F.claude\u002Fskills\u002Foptimize-skills\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: optimize-skills\ndescription: Compile your local stellar-build usage traces into sharper skills. Use when the user says \"optimize my skills\", \"optimize-skills\", \"improve my skills from usage\", \"learn from my traces\", \"stellar-loop optimize\", or wants the learning loop to refine installed skills based on how they've actually been used. DSPy-style prompt optimization over the local trace store at ~\u002F.stellar\u002Ftraces - fully local, reversible.\n---\n\n## What this skill does\n\nThis is the **optimizer** half of the stellar-build learning loop - the analog of\n`jarvis optimize skills --policy dspy`. It reads the local trace of how you've\nactually used your skills and *compiles* that experience back into the installed\nskill files: refined trigger descriptions, tightened instructions, and mined\nfew-shot examples drawn from your own successful runs.\n\nThink of it in DSPy terms:\n- **Trainset** = your trace records in `~\u002F.stellar\u002Ftraces\u002F*.jsonl`, grouped into\n  **spans** (one continuous use of a skill within a session - see below)\n- **Metric** = an LLM-as-judge quality label you assign to each *span*\n- **Compile** = rewriting the skill's instructions + few-shot demos from the wins\n\nEverything is local and reversible. The original of every edited skill is backed\nup to `~\u002F.stellar\u002Fbackups\u002F\u003Cskill>\u002FSKILL.md.\u003Ctimestamp>` before any change, and\n`stellar-loop restore \u003Cskill>` undoes the last optimization.\n\n## How to run it\n\nDo these steps in order. Be conservative: this edits files the user relies on.\n\n### 1. Load the traces and group them into spans\n\nRead the trace store:\n\n```\n~\u002F.stellar\u002Ftraces\u002F*.jsonl\n```\n\nA skill *use* is a multi-turn **span**, not a single tool call, so records come\nin three shapes - key off the `kind` field:\n\n- `kind:\"turn\"` - one user→assistant turn while a skill was resident:\n  `{ kind:\"turn\", skill, prompt, result_preview, result_len, span_id, turn_index, session, ts }`.\n  `result_preview` is the **real assistant output** for that turn (capped ~1500 chars).\n- `kind:\"span\"` - the summary written when the span closed (at SessionEnd or on a\n  persona handoff): `{ kind:\"span\", skill, prompt (opening), result_preview (final\n  work product), turns, span_id, session, ts }`.\n- **Legacy \u002F coverage-only** - older records and live Codex records have **no\n  `kind`** and `event:\"skill_invoked\"`; their `result_preview` is just the mount\n  acknowledgment (`{\"success\":true,\"commandName\":\"…\"}`, ~47 chars). These count\n  toward coverage but carry **no work product** - never mine demos from them.\n\n**Group records by span**: a span is identified by `span_id` (fall back to\n`session` × `skill` for legacy records). One span = one usage, regardless of how\nmany turns it spans. The `kind:\"span\"` record is the span's summary; the\n`kind:\"turn\"` records are its detail.\n\nIf `STELLAR_HOME` is set in the environment, use that instead of `~\u002F.stellar`.\nIf there are no traces, tell the user to use some skills first (and confirm the\nhooks are registered - `stellar-loop status` shows this), then stop.\n\nIf the user named a specific skill (e.g. \"optimize find-stellar-idea\"), filter to\nthat skill. Otherwise consider every skill with **at least 5 spans** - fewer than\nthat is too little signal to compile from; say so and skip it. (Count spans, not\nraw records: five turns of one resident-skill session is **one** span, not five.)\n\n### 2. Score each trace (the metric)\n\nFor each candidate skill, judge each **span** (not each raw record) as an honest,\nskeptical evaluator. Label each span **win \u002F neutral \u002F loss** using signals you\ncan see:\n- Did the opening `prompt` \u002F `args` clearly match what the skill is for? (good\n  trigger fit)\n- Does the work product - the `kind:\"span\"` `result_preview`, or the\n  concatenated `kind:\"turn\"` previews - show the skill did its job vs. flailing,\n  asking for re-clarification, or producing something off-target?\n- **Re-prompts within a span** are the loss signal: several `kind:\"turn\"` records\n  under one `span_id` where the user keeps re-asking or correcting (rather than\n  building forward) suggest the skill missed on the first pass. Judge by the\n  *turns within the span*, not by counting separate activations - a resident\n  skill legitimately spans many turns.\n- A long, on-topic work product after a precise opening prompt is a win signal.\n\nBe calibrated, not generous. If you're unsure, call it neutral. Keep only the\n**winning spans** as material to learn from.\n\n### 3. Compile (the optimization)\n\nFor each skill, read its currently installed file:\n\n```\n~\u002F.claude\u002Fskills\u002F\u003Cskill>\u002FSKILL.md\n```\n\nNow produce an improved version. Two levers, exactly like DSPy optimizes both\ninstructions and demonstrations:\n\n**(a) Instructions & trigger.** If the wins reveal that the user consistently uses\nthis skill for a narrower or different purpose than the description claims, tighten\nthe `description` triggers and the body instructions to match reality. Add defaults\nthe user clearly prefers (e.g. they always want the weekend-hack framing, or always\ntarget testnet first). Never delete the skill's core capability - only sharpen it.\n\n**(b) Few-shot demos.** Distill 1–3 of the strongest **winning spans** into compact,\ngeneralized examples. Mine them from `kind:\"span\"` \u002F `kind:\"turn\"` records - these\nnow hold the real assistant work product (the opening prompt paired with what the\nskill actually produced), which is exactly the demonstration material DSPy needs.\n**Never mine a demo from a legacy \u002F coverage-only record** (no `kind`, mount-stub\n`result_preview`): it has no work product to learn from. Place the demos in a\nclearly delimited, regenerable block at the END of the skill body:\n\n```markdown\n\u003C!-- LEARNED:BEGIN (auto-generated by optimize-skills - safe to delete\u002Fregenerate) -->\n## Learned from your usage\n\n_Compiled from \u003CN> local traces on \u003CUTC date>. These are personalized defaults and\nexamples mined from how you actually use this skill. Edit or delete freely; rerun\n`\u002Foptimize-skills` to regenerate. Undo with `stellar-loop restore \u003Cskill>`._\n\n**Defaults you tend to want:** \u003Cshort bullet list, or omit if none clear>\n\n**Example interactions that worked well:**\n- _When the user asked_ \"\u003Cgeneralized trigger>\", _the effective move was_ \"\u003Cdistilled approach>\".\n- ...\n\u003C!-- LEARNED:END -->\n```\n\nRules for the learned block:\n- It is the ONLY part of the file the optimizer owns. Replace any existing\n  `LEARNED:BEGIN..END` block wholesale; never stack duplicates.\n- Keep it short (aim \u003C 40 lines). Generalize - never paste raw user data, secrets,\n  paths, addresses, or anything that looks sensitive from `result_preview`.\n- Preserve the YAML frontmatter `name` exactly. You may sharpen `description`.\n\n### 4. Back up, then write\n\nBefore writing any skill, copy the current file to a timestamped backup:\n\n```\nmkdir -p ~\u002F.stellar\u002Fbackups\u002F\u003Cskill>\ncp ~\u002F.claude\u002Fskills\u002F\u003Cskill>\u002FSKILL.md ~\u002F.stellar\u002Fbackups\u002F\u003Cskill>\u002FSKILL.md.\u003CUTC-timestamp>\n```\n\nThen write the improved `SKILL.md` to BOTH install locations if they exist:\n- `~\u002F.claude\u002Fskills\u002F\u003Cskill>\u002FSKILL.md`\n- `~\u002F.codex\u002Fskills\u002F\u003Cskill>\u002FSKILL.md`\n\n### 5. Refresh the learned profile (feeds the auto-reprompter)\n\nAcross ALL the traces (not per-skill), distill a short profile of how this user\nworks and write it to `~\u002F.stellar\u002Fprofile.md` (honor `STELLAR_HOME`). The\nper-turn auto-reprompter injects this into every prompt, so it's how the loop\nmakes your *prompts* - not just your skills - get sharper over time.\n\nKeep it to ~6–10 tight bullets of durable, generalizable signal, e.g.:\n- domain(s) they keep returning to (payments, DeFi\u002FSoroban, RWA, …)\n- typical scope\u002Ftimeline (weekend hack vs. quarter-long)\n- stack \u002F tools they favor, environment defaults (e.g. testnet-first)\n- recurring preferences in how they want answers (brevity, code-first, persona)\n\nRules: generalize only - no secrets, addresses, file paths, or raw snippets.\nOverwrite the file each run (it's a living profile, not a log). If the signal is\ntoo thin to say anything useful, write nothing rather than guessing.\n\n### 6. Record the optimization\n\nAppend one line per optimized skill to `~\u002F.stellar\u002Foptimizations.jsonl`:\n\n```json\n{\"ts\":\"\u003CUTC>\",\"event\":\"optimized\",\"skill\":\"\u003Cname>\",\"traces_seen\":\u003Cn>,\"wins\":\u003Cn>,\"changed\":[\"description\",\"learned-block\"],\"backup\":\"\u003Cpath>\"}\n```\n\n### 7. Report\n\nSummarize what changed: which skills, how many traces\u002Fwins fed each, a one-line\ndiff-in-words per skill, and the exact `stellar-loop restore \u003Cskill>` command to\nundo. Recommend `\u002Fbench-skills` to measure whether the change actually helped.\n\n## Guardrails\n\n- **Reversible always.** No edit without a fresh backup first.\n- **Local only.** Read only from `~\u002F.stellar` and the installed skills. Never send\n  traces anywhere.\n- **Conservative.** When the signal is thin or ambiguous, leave the skill unchanged\n  and say why. A skill that doesn't improve should not be touched.\n- **No sensitive data in skills.** Generalize every mined example.\n- **Don't optimize the loop's own skills** (`optimize-skills`, `bench-skills`,\n  `learning-loop`) - skip them to avoid feedback loops.\n","## What this skill does\n\nThis is the **optimizer** half of the stellar-build learning loop - the analog of\n`jarvis optimize skills --policy dspy`. It reads the local trace of how you've\nactually used your skills and *compiles* that experience back into the installed\nskill files: refined trigger descriptions, tightened instructions, and mined\nfew-shot examples drawn from your own successful runs.\n\nThink of it in DSPy terms:\n- **Trainset** = your trace records in `~\u002F.stellar\u002Ftraces\u002F*.jsonl`, grouped into\n  **spans** (one continuous use of a skill within a session - see below)\n- **Metric** = an LLM-as-judge quality label you assign to each *span*\n- **Compile** = rewriting the skill's instructions + few-shot demos from the wins\n\nEverything is local and reversible. The original of every edited skill is backed\nup to `~\u002F.stellar\u002Fbackups\u002F\u003Cskill>\u002FSKILL.md.\u003Ctimestamp>` before any change, and\n`stellar-loop restore \u003Cskill>` undoes the last optimization.\n\n## How to run it\n\nDo these steps in order. Be conservative: this edits files the user relies on.\n\n### 1. Load the traces and group them into spans\n\nRead the trace store:\n\n```\n~\u002F.stellar\u002Ftraces\u002F*.jsonl\n```\n\nA skill *use* is a multi-turn **span**, not a single tool call, so records come\nin three shapes - key off the `kind` field:\n\n- `kind:\"turn\"` - one user→assistant turn while a skill was resident:\n  `{ kind:\"turn\", skill, prompt, result_preview, result_len, span_id, turn_index, session, ts }`.\n  `result_preview` is the **real assistant output** for that turn (capped ~1500 chars).\n- `kind:\"span\"` - the summary written when the span closed (at SessionEnd or on a\n  persona handoff): `{ kind:\"span\", skill, prompt (opening), result_preview (final\n  work product), turns, span_id, session, ts }`.\n- **Legacy \u002F coverage-only** - older records and live Codex records have **no\n  `kind`** and `event:\"skill_invoked\"`; their `result_preview` is just the mount\n  acknowledgment (`{\"success\":true,\"commandName\":\"…\"}`, ~47 chars). These count\n  toward coverage but carry **no work product** - never mine demos from them.\n\n**Group records by span**: a span is identified by `span_id` (fall back to\n`session` × `skill` for legacy records). One span = one usage, regardless of how\nmany turns it spans. The `kind:\"span\"` record is the span's summary; the\n`kind:\"turn\"` records are its detail.\n\nIf `STELLAR_HOME` is set in the environment, use that instead of `~\u002F.stellar`.\nIf there are no traces, tell the user to use some skills first (and confirm the\nhooks are registered - `stellar-loop status` shows this), then stop.\n\nIf the user named a specific skill (e.g. \"optimize find-stellar-idea\"), filter to\nthat skill. Otherwise consider every skill with **at least 5 spans** - fewer than\nthat is too little signal to compile from; say so and skip it. (Count spans, not\nraw records: five turns of one resident-skill session is **one** span, not five.)\n\n### 2. Score each trace (the metric)\n\nFor each candidate skill, judge each **span** (not each raw record) as an honest,\nskeptical evaluator. Label each span **win \u002F neutral \u002F loss** using signals you\ncan see:\n- Did the opening `prompt` \u002F `args` clearly match what the skill is for? (good\n  trigger fit)\n- Does the work product - the `kind:\"span\"` `result_preview`, or the\n  concatenated `kind:\"turn\"` previews - show the skill did its job vs. flailing,\n  asking for re-clarification, or producing something off-target?\n- **Re-prompts within a span** are the loss signal: several `kind:\"turn\"` records\n  under one `span_id` where the user keeps re-asking or correcting (rather than\n  building forward) suggest the skill missed on the first pass. Judge by the\n  *turns within the span*, not by counting separate activations - a resident\n  skill legitimately spans many turns.\n- A long, on-topic work product after a precise opening prompt is a win signal.\n\nBe calibrated, not generous. If you're unsure, call it neutral. Keep only the\n**winning spans** as material to learn from.\n\n### 3. Compile (the optimization)\n\nFor each skill, read its currently installed file:\n\n```\n~\u002F.claude\u002Fskills\u002F\u003Cskill>\u002FSKILL.md\n```\n\nNow produce an improved version. Two levers, exactly like DSPy optimizes both\ninstructions and demonstrations:\n\n**(a) Instructions & trigger.** If the wins reveal that the user consistently uses\nthis skill for a narrower or different purpose than the description claims, tighten\nthe `description` triggers and the body instructions to match reality. Add defaults\nthe user clearly prefers (e.g. they always want the weekend-hack framing, or always\ntarget testnet first). Never delete the skill's core capability - only sharpen it.\n\n**(b) Few-shot demos.** Distill 1–3 of the strongest **winning spans** into compact,\ngeneralized examples. Mine them from `kind:\"span\"` \u002F `kind:\"turn\"` records - these\nnow hold the real assistant work product (the opening prompt paired with what the\nskill actually produced), which is exactly the demonstration material DSPy needs.\n**Never mine a demo from a legacy \u002F coverage-only record** (no `kind`, mount-stub\n`result_preview`): it has no work product to learn from. Place the demos in a\nclearly delimited, regenerable block at the END of the skill body:\n\n```markdown\n\u003C!-- LEARNED:BEGIN (auto-generated by optimize-skills - safe to delete\u002Fregenerate) -->\n## Learned from your usage\n\n_Compiled from \u003CN> local traces on \u003CUTC date>. These are personalized defaults and\nexamples mined from how you actually use this skill. Edit or delete freely; rerun\n`\u002Foptimize-skills` to regenerate. Undo with `stellar-loop restore \u003Cskill>`._\n\n**Defaults you tend to want:** \u003Cshort bullet list, or omit if none clear>\n\n**Example interactions that worked well:**\n- _When the user asked_ \"\u003Cgeneralized trigger>\", _the effective move was_ \"\u003Cdistilled approach>\".\n- ...\n\u003C!-- LEARNED:END -->\n```\n\nRules for the learned block:\n- It is the ONLY part of the file the optimizer owns. Replace any existing\n  `LEARNED:BEGIN..END` block wholesale; never stack duplicates.\n- Keep it short (aim \u003C 40 lines). Generalize - never paste raw user data, secrets,\n  paths, addresses, or anything that looks sensitive from `result_preview`.\n- Preserve the YAML frontmatter `name` exactly. You may sharpen `description`.\n\n### 4. Back up, then write\n\nBefore writing any skill, copy the current file to a timestamped backup:\n\n```\nmkdir -p ~\u002F.stellar\u002Fbackups\u002F\u003Cskill>\ncp ~\u002F.claude\u002Fskills\u002F\u003Cskill>\u002FSKILL.md ~\u002F.stellar\u002Fbackups\u002F\u003Cskill>\u002FSKILL.md.\u003CUTC-timestamp>\n```\n\nThen write the improved `SKILL.md` to BOTH install locations if they exist:\n- `~\u002F.claude\u002Fskills\u002F\u003Cskill>\u002FSKILL.md`\n- `~\u002F.codex\u002Fskills\u002F\u003Cskill>\u002FSKILL.md`\n\n### 5. Refresh the learned profile (feeds the auto-reprompter)\n\nAcross ALL the traces (not per-skill), distill a short profile of how this user\nworks and write it to `~\u002F.stellar\u002Fprofile.md` (honor `STELLAR_HOME`). The\nper-turn auto-reprompter injects this into every prompt, so it's how the loop\nmakes your *prompts* - not just your skills - get sharper over time.\n\nKeep it to ~6–10 tight bullets of durable, generalizable signal, e.g.:\n- domain(s) they keep returning to (payments, DeFi\u002FSoroban, RWA, …)\n- typical scope\u002Ftimeline (weekend hack vs. quarter-long)\n- stack \u002F tools they favor, environment defaults (e.g. testnet-first)\n- recurring preferences in how they want answers (brevity, code-first, persona)\n\nRules: generalize only - no secrets, addresses, file paths, or raw snippets.\nOverwrite the file each run (it's a living profile, not a log). If the signal is\ntoo thin to say anything useful, write nothing rather than guessing.\n\n### 6. Record the optimization\n\nAppend one line per optimized skill to `~\u002F.stellar\u002Foptimizations.jsonl`:\n\n```json\n{\"ts\":\"\u003CUTC>\",\"event\":\"optimized\",\"skill\":\"\u003Cname>\",\"traces_seen\":\u003Cn>,\"wins\":\u003Cn>,\"changed\":[\"description\",\"learned-block\"],\"backup\":\"\u003Cpath>\"}\n```\n\n### 7. Report\n\nSummarize what changed: which skills, how many traces\u002Fwins fed each, a one-line\ndiff-in-words per skill, and the exact `stellar-loop restore \u003Cskill>` command to\nundo. Recommend `\u002Fbench-skills` to measure whether the change actually helped.\n\n## Guardrails\n\n- **Reversible always.** No edit without a fresh backup first.\n- **Local only.** Read only from `~\u002F.stellar` and the installed skills. Never send\n  traces anywhere.\n- **Conservative.** When the signal is thin or ambiguous, leave the skill unchanged\n  and say why. A skill that doesn't improve should not be touched.\n- **No sensitive data in skills.** Generalize every mined example.\n- **Don't optimize the loop's own skills** (`optimize-skills`, `bench-skills`,\n  `learning-loop`) - skip them to avoid feedback loops.\n",{"name":5,"description":7},"main","db2f47e2b79822ad3d8dd0f17489d5d15bcba96d",".claude\u002Fskills\u002Foptimize-skills",3,"https:\u002F\u002Fgithub.com\u002FWritzProtocol\u002Fwritz","https:\u002F\u002Fwritz.xyz","https:\u002F\u002Fgithub.com\u002FWritzProtocol","https:\u002F\u002Fwritz.xyz\u002F","Native Bitcoin lending on Stellar. No bridge, no custodian.","2026-11-14 18:13:11",{"source":8,"skillName":5,"commands":35,"prompt":45,"zipAvailable":46},[36,39,42],{"tool":37,"command":38},"npx","npx skills add WritzProtocol\u002Fwritz --skill optimize-skills",{"tool":40,"command":41},"bunx","bunx skills add WritzProtocol\u002Fwritz --skill optimize-skills",{"tool":43,"command":44},"pnpm","pnpm dlx skills add WritzProtocol\u002Fwritz --skill optimize-skills","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FWritzProtocol\u002Fwritz\u002Fblob\u002Fmain\u002F.claude\u002Fskills\u002Foptimize-skills\u002FSKILL.md",true,[],[49,56,63,70,77,84],{"skillNo":50,"name":51,"nameZh":6,"description":52,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":53,"categories":54,"keywords":55,"publishedAt":19},"01M77KKJ6DYZP10BJ74CW4H0AE","zk-proofs","Zero-knowledge proofs and privacy patterns on Stellar. Covers Groth16 verification in smart contracts via BLS12-381 host functions (CAP-0059, available), the BN254 + Poseidon proposals (CAP-0074\u002F0075, status-sensitive), and concrete toolchain walkthroughs for Circom (on-chain verifiable today), Noir, and RISC Zero (attestation pattern until BN254 lands). Use when building privacy-preserving applications, ZK-verifier contracts, or wiring a proving toolchain to Stellar.","https:\u002F\u002Fgithub.com\u002FWritzProtocol\u002Fwritz\u002Fblob\u002Fmain\u002F.codex\u002Fskills\u002Fzk-proofs\u002FSKILL.md",[],[],{"skillNo":57,"name":58,"nameZh":6,"description":59,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":60,"categories":61,"keywords":62,"publishedAt":19},"01M77KKF3SK8JVBGRY4A89HYQC","tyler-architect","System architect and technical design leader. Use when the user asks to talk to Tyler or requests the architect.","https:\u002F\u002Fgithub.com\u002FWritzProtocol\u002Fwritz\u002Fblob\u002Fmain\u002F.codex\u002Fskills\u002Ftyler-architect\u002FSKILL.md",[],[],{"skillNo":64,"name":65,"nameZh":6,"description":66,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":67,"categories":68,"keywords":69,"publishedAt":19},"01M77KKCD584VQA1TM1F9Q8JBH","stellar-help","Analyzes current state and user query to answer BMad questions or recommend the next skill(s) to use. Use when user asks for help, bmad help, what to do next, or what to start with in BMad.","https:\u002F\u002Fgithub.com\u002FWritzProtocol\u002Fwritz\u002Fblob\u002Fmain\u002F.codex\u002Fskills\u002Fstellar-help\u002FSKILL.md",[],[],{"skillNo":71,"name":72,"nameZh":6,"description":73,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":74,"categories":75,"keywords":76,"publishedAt":19},"01M77KK7W2HKQ3QMAG0KJY1BHH","stellar-competitive-landscape","Map the competitive landscape for a Stellar project idea. Use when a user says \"who are my competitors on Stellar\", \"competitive analysis Stellar\", \"what already exists in this space on Stellar\", \"show me similar Stellar projects\", \"is anyone else building this on Stellar\", or \"map the Stellar landscape for X\". Queries the 728-project LumenLoop ecosystem database and Electric Capital developer activity to rank competitors by SCF funding history and repo activity.","https:\u002F\u002Fgithub.com\u002FWritzProtocol\u002Fwritz\u002Fblob\u002Fmain\u002F.codex\u002Fskills\u002Fstellar-competitive-landscape\u002FSKILL.md",[],[],{"skillNo":78,"name":79,"nameZh":6,"description":80,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":81,"categories":82,"keywords":83,"publishedAt":19},"01M77KK1ACXYN5AZTWR8B5CKTX","standards","Stellar standards, ecosystem, and reference. Covers SEPs (Stellar Ecosystem Proposals), CAPs (Core Advancement Proposals), and a quick map for picking the right standard for wallets, anchors, payments, deposits\u002Fwithdrawals, federation, deep links, and KYC. Also bundles ecosystem references (DeFi protocols, dev tools, wallets, infra, community projects), curated documentation links, and MCP servers (live tools such as Raven). Use when you need to know which SEP applies, or want a starting point for ecosystem integrations, official docs, or live MCP tooling.","https:\u002F\u002Fgithub.com\u002FWritzProtocol\u002Fwritz\u002Fblob\u002Fmain\u002F.codex\u002Fskills\u002Fstandards\u002FSKILL.md",[],[],{"skillNo":85,"name":86,"nameZh":6,"description":87,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":88,"categories":89,"keywords":90,"publishedAt":19},"01M77KJV8VW46WC58DHJR2296C","smart-contracts","Stellar smart contract development (Rust, soroban-sdk). Entry point with project setup, contract anatomy, and build\u002Fdeploy workflow, routing to three companion files in this directory - development.md (storage\u002FTTL, authorization, cross-contract calls, tokens, events, errors, upgrades, fees, troubleshooting), testing.md (unit, fuzz, property, fork, mutation, integration), and security.md (vulnerability classes, checklists, tooling, audits). Use when writing, testing, reviewing, securing, debugging, or shipping Stellar smart contracts, including anything the user calls \"Soroban\" - Soroban contracts, soroban-sdk, Soroban auth\u002Fstorage\u002FTTL errors, SEP-41 tokens, or SAC integration from contract code.","https:\u002F\u002Fgithub.com\u002FWritzProtocol\u002Fwritz\u002Fblob\u002Fmain\u002F.codex\u002Fskills\u002Fsmart-contracts\u002FSKILL.md",[],[]]