संदर्भ
स्किल
SKILL.md फ़ाइलों में पुन: प्रयोज्य पता-योग्य तकनीकें, प्लान-सूट और परितंत्र-ऑडिट के लिए पहचान संकेतों, प्रक्रियाओं, और प्रवर्तन हुक के साथ।
स्किल src/apothem/skills/ के अंतर्गत फ़ोल्डरों के रूप में रहती हैं, प्रत्येक में एक SKILL.md प्रवेश
बिंदु। स्किल पुन: प्रयोज्य पता-योग्य तकनीकों को पकड़ती हैं — उनमें एक
पहचान संकेत, एक प्रक्रिया, और एक प्रवर्तन हुक होता है।
संपूर्ण स्किल सूची इस पृष्ठ के निचले हिस्से में «उत्पन्न स्किल सूची» खंड में उत्पन्न होती है।
Frontmatter कॉन्ट्रैक्ट
SKILL.md धारण करती है:
---
name: <skill name>
version: <semver>
updated: <YYYY-MM-DD>
description: <one-line role>
archetype: <class-template, e.g. workflow-template>
userInvocable: <true | false>
disable-model-invocation: <true | false>
allowed-tools: <comma-separated tool list, or "*">
---लेखन अनुशासन
- प्रति-स्किल फ़ोल्डर। हर स्किल अपना स्वयं का फ़ोल्डर है, जिसका प्रवेश
बिंदु
SKILL.mdपर है। - पहचान संकेत। स्किल बॉडी उस पहचान संकेत के साथ खुलती हैं जो तकनीक को ट्रिगर करता है।
- प्रक्रिया + प्रवर्तन। स्किल दोनों धारण करती हैं — लागू करने की प्रक्रिया और लूप को बंद करने वाला प्रवर्तन संकेत।
उत्पन्न स्किल सूची
| Name | Description | Source |
|---|---|---|
dependency-upgrade | Changelog-reviewed, pinned, gate-verified dependency upgrade — matched when the operator asks to 'bump deps', 'upgrade dependencies', 'update packages', 'bump the lockfile', 'refresh dependency versions', or any phrasing requesting an audited version bump of the host project's third-party dependencies. Five-step loop: enumerate outdated dependencies from the host's manifest + lockfile, classify each candidate against its changelog (safe / breaking / major), apply pinned bumps honoring the host's discovered pin policy, run the host's lint / test / type-check gates, and report. Breaking and major candidates STOP for structured inquiry — never crossed silently. Idempotent (re-run on a current manifest yields zero bumps); --dry-run classifies without writing; --package NAME scopes to one dependency. Developer + security cohorts; user-invocable; honors host-discovered tooling and assumes no package manager. | src/apothem/skills/dependency-upgrade/SKILL.md |
dev-toolkit | Engineering-discipline toolkit — matched when the operator asks to 'diagnose a bug', 'debug systematically', 'do TDD / write a failing test first', 'break this into vertical slices', 'turn this into a PRD / issues', 'decompose this work', or otherwise needs a disciplined software-engineering loop rather than an ad-hoc fix. Sequences four named, gated, composable loops — structured debugging (reproduce → minimize → hypothesize → instrument → fix root cause), test-first red-green-refactor (red observed before green), vertical-slice decomposition (independently-shippable testable units), and requirement-to-issue breakdown (goal → testable requirement set → tracked issues + ADR) — selected by --mode or by the request. Cross-references (never duplicates) sibling owners: orchestration → /workflow, diagrams → diagram-authoring, minimal-diff guarding → surgical-guard. Harness-agnostic; deterministic output; drives the host's discovered toolchain. | src/apothem/skills/dev-toolkit/SKILL.md |
diagram-authoring | Diagram authoring and rendering — matched when the operator asks to 'draw a diagram', 'make a flowchart / sequence / state / ER / architecture diagram', 'author a data visualization', 'render this Mermaid', 'export a diagram to SVG/PNG/PDF', or otherwise needs a structural picture authored and rendered. One deterministic pipeline — pick notation by subject → author source → validate syntax → render → export — spanning declarative diagrams (flowchart / sequence / state / class / ER / gantt / mindmap) and data-driven visualizations (bar / line / scatter / network / hierarchy / geo). Correct-by-construction (validated before render) and portable (SVG / PNG / PDF via a discovered headless renderer). Not for: shipping or assuming a vendor render server (the renderer is discovered, never bundled); fabricating a visualization's dataset (supplied or discovered, never invented); illustrating non-structural prose a paragraph already carries. Harness-agnostic; deterministic output. | src/apothem/skills/diagram-authoring/SKILL.md |
document-authoring | Long-form structured-document authoring — matched when the operator asks to 'write a thesis / dissertation / paper / report', 'author a LaTeX document', 'build a document from an outline', 'compile this to PDF', 'manage citations / bibliography', or otherwise needs a long, citation-bearing, typeset document produced. Drives an approval-gated hierarchical pipeline (outline → operator-approve → draft → review), verifies every citation against a real retrievable source (never fabricated — unresolvable references land in an unverified-citations ledger), and compiles through a deterministic detect→route pipeline that inspects the source, selects engine / bibliography tool / index passes by rule, and self-corrects one recoverable build error before surfacing failure. Figures route to diagram-authoring. Not for: a short prose snippet or single-paragraph rewrite (ordinary prose authoring, below threshold); fabricating a citation to fill a gap; rendering figures. Harness-agnostic; deterministic output. | src/apothem/skills/document-authoring/SKILL.md |
ecosystem-audit | Blind audit of the apothem ecosystem — matched when the user asks to 'double check', 'audit', 'verify ecosystem', 'sweep', 'validate ecosystem', 'review configuration', 'check the apothem tree', or any phrasing requesting a comprehensive validation of the apothem configuration tree as a whole. Five-phase cadence (census → parallel cross-reference audit → synthesis → report → optional --fix) detects drift, staleness, orphans, dangling references, conflicting directives, frontmatter invalidity, registry-vs-disk count mismatches, and secret exposure. Findings classify FIX / ENHANCE / CONSIDER / DEFER; only FIX auto-applies under --fix, and only after a per-file destructive-op confirmation — ENHANCE / CONSIDER / DEFER surface for operator decision via the structured-inquiry channel. User-invocable directly; also dispatchable from /plan-execute discovery contexts. | src/apothem/skills/ecosystem-audit/SKILL.md |
eval-harness | Build and run a reproducible LLM evaluation harness — matched when the operator asks to 'build an eval', 'measure the model', set up an 'evaluation harness', 'score outputs', benchmark a prompt or model variant, regression-test generation quality, or compare candidate models on a labeled task. Defines four artifacts: a versioned labeled dataset, a scorer (exact-match / rubric / LLM-judge) with an explicit pass criterion, a candidate runner, and an aggregated metrics report carrying pass-rate with a Wilson confidence interval, per-category breakdowns, and regression/improvement lists against a prior run. Reproducibility is the contract — every run stamps the dataset version, the scorer definition, the candidate identifier, and the seed so two runs over identical inputs return an identical verdict. Harness- and model-agnostic: drives any provider through its native invocation surface, discovered or inquired per rules/host-discovery.md. NOT for fine-tuning or training pipelines, authoring the prompt under test, live-monitoring dashboards, one-shot un-versioned benchmarks, or scoring against an undefined pass criterion. User-invocable directly. | src/apothem/skills/eval-harness/SKILL.md |
incident-runbook | Author and execute operational incident runbooks — matched when the operator asks to 'write a runbook', define an 'on-call procedure', author 'incident response steps', document a 'recovery procedure', 'document recovery steps for a failing service', or capture the response sequence for a failing or compromised service. Produces a runbook in the canonical five-stage shape — Trigger → Diagnosis → Action → Verification → Rollback — where every Action step carries an explicit, paired rollback path and every irreversible Action sits behind a no-default confirmation gate. On execution, advances Diagnosis → Action → Verification in order, confirming each live-infrastructure Action through the structured-inquiry channel and STOPping at the first Verification failure to surface its rollback. Serves the developer and security cohorts. NOT for provisioning monitoring/alerting/dashboards, paging on-call rotations or opening incident channels, writing post-mortems or retrospectives, or batch-executing destructive remediation unattended. User-invocable directly. | src/apothem/skills/incident-runbook/SKILL.md |
multi-source-research | Multi-source research harness — matched when the operator asks to 'research X deeply', 'investigate', 'find sources on', 'gather evidence about', 'fact-check', or any phrasing that demands a fanned-out, cross-verified, cited report rather than a single answer recalled from memory. Runs a five-step pipeline: decompose the question into testable sub-questions, fan out parallel source-discovery queries through a Research Team agent dispatch, fetch and extract claims per source, adversarially verify each claim against two or more independent sources, and synthesize a cited report where every claim carries its sources, a HIGH/MEDIUM/LOW confidence level, and an open-questions list. A single-source claim never closes at HIGH; an unverifiable claim is reported at LOW with the gap named, never papered over with an invented citation. User-invocable directly. NOT a single-answer lookup, a code generator, a plan-suite generator, a documentation generator, or a single-source summarizer. | src/apothem/skills/multi-source-research/SKILL.md |
plan-suite | Master Plan Suite template container — matched when the work is structured plan generation from authored prose, phase-by-phase decomposition of a complex multi-step engagement, plan-suite refinement, forensic plan review, closed-loop plan audit/remediation, architectural design, phase execution against quality gates, read-only progress reporting, or decision-preserving amendment; consumed by the /plan pipeline stages (spec, generate, review, design, audit, execute, status, amend). Houses master-template.md — the canonical specification defining plan-suite structure, the mandate catalog (TM-1–28), the core-principle catalog (CP-1–27), and the Technical Co-Founder Framework. Not directly user-invocable: the /plan stages resolve the template by path, cold-load it on a context-empty invocation, and refuse to regenerate it from memory if it is missing. Not a codegen tool, not a research pipeline, not a documentation generator, not a registry, and not stateful across sessions. | src/apothem/skills/plan-suite/SKILL.md |
projectify | Chat-app Project elevation skill — matched when the operator asks to 'projectify', 'build a Project', 'set up a Custom GPT / Gem', 'write Project instructions', 'rewrite my Project', or otherwise hands off a chat-app Project (Claude Project / ChatGPT Custom GPT / Gemini Gem) to be freshly authored or elevated to current SOTA conventions. Emits the three deliverables a Project carries — Description, Instruction, knowledge Files — through an inquiry-saturated elicitation that reconciles every preference and ambiguity with the operator before committing, elevated across order / coherence / clarity / determinism / structurality / conciseness / rigor / comprehensiveness and beyond. Holds the knowledge-file set within a measurable per-platform context budget (token-sum / platform-limit <= 0.02, the limit discovered live and never inlined), with the grant to consolidate or divide files through the inquiry channel. Installs agnostically across all harnesses; the deliverables target the chat-app Project. Deterministic output; multi-step autonomy is opt-in / confirmation-gated. NOT for authoring coding-harness rules / skills / commands / hooks (route to the ecosystem authoring paths), and NOT a hardcoded-limit calculator (the per-platform limit is discovered live, never inlined). | src/apothem/skills/projectify/SKILL.md |
prompt-engineering | Design and eval-validate structured prompts for language-model systems — matched when the user asks to 'design a prompt', 'write a system prompt for', 'improve this prompt', 'engineer a prompt', 'build a prompt template', or any phrasing that asks for a model-driven instruction surface backed by falsifiable evaluation. Clarifies the task contract, drafts a structured prompt with explicit role / instructions / output schema, defines happy / edge / adversarial eval cases, iterates against them, and records the prompt with its eval rationale. Produces a prompt plus its eval set — NOT a running agent, a fine-tuning pipeline, a cross-model benchmark, or a model-selection decision. Model-agnostic and harness-agnostic; no single vendor is assumed. User-invocable directly. | src/apothem/skills/prompt-engineering/SKILL.md |
refactor-extract | Scoped behavior-preserving extraction refactor — matched when the user asks to 'extract this', 'extract this function', 'pull this out into its own module', 'clean up this module', 'refactor this', 'split this function', 'break this apart', or any phrasing that asks for a structural extraction of an existing symbol or file while preserving observable behavior. Performs behavioral extraction of what the code DOES, clean-room re-derivation from the extracted specification, quality elevation against a named deficiency, and regression verification that behavior is preserved. NOT for behavior changes (changed inputs / outputs / side effects / error behavior route back as a feature-change finding), whole-file rewrites (the clean-room barrier binds the named target only), or building a test suite as the deliverable. User-invocable directly with an optional target symbol or file — the interactive skill counterpart to the dispatched refactor-surgeon agent. | src/apothem/skills/refactor-extract/SKILL.md |
research-suite | Research Suite template — matched when the request is a structured research engagement from problem ideation and question formulation, conceptual / theoretical framing, evidence-gathering across primary sources, literature synthesis with an explicit gap statement, proposal authoring, falsifiable study design and preregistration, reproducible experiment execution, preregistered statistical analysis, paper authoring, peer-review-grade adversarial review, venue-formatted publication, or post-acceptance dissemination and impact; consumed by the /research pipeline commands (ideate, spec, theory, sources, synthesis, proposal, design, experiment, analysis, paper, review, publish, disseminate). Houses the canonical surface that defines research-suite structure, the ten rigor mandates (R1–R10), the thirteen-stage research lifecycle, and the Principal-Investigator Framework. Not directly user-invocable; the consuming /research pipeline stages resolve the surface by path. | src/apothem/skills/research-suite/SKILL.md |
secret-rotation | Safe secret-rotation template — matched when the user says 'rotate a secret', 'a key leaked', 'a token leaked', 'a credential is exposed', 'revoke and re-issue this key', 'an API key was committed', or any phrasing that asks to safely retire and replace an exposed credential. Treats a leaked credential as compromised the moment it touched a tracked file and enforces the revoke-before-re-issue ordering invariant: detects the coarse exposure surface, revokes the compromised credential at its issuer FIRST, re-issues a least-privilege replacement, rewires configuration through env-var or secret-manager indirection (never plaintext), and verifies no plaintext residue remains in the working tree or version-control history. NOT a full secret-scanner (routes the broad sweep to the host's scanner), NOT an issuer client, NOT a history-rewrite engine. User-invocable directly. | src/apothem/skills/secret-rotation/SKILL.md |
source-synthesis | Reconcile a fixed set of provided sources into one cited synthesis — matched when the user says 'synthesize these sources', 'combine these documents', 'reconcile these', 'what do these say together', 'merge these notes into one', or any phrasing that hands over N specific documents and asks for a single consolidated reading. Extracts the claim set per source with citation anchors, reconciles agreements and contradictions across sources, and produces a cited synthesis that separates consensus from contested ground while flagging coverage gaps. User-invocable directly via --sources; never gathers or discovers sources of its own (that is the multi-source-research surface), never adjudicates which source is correct against ground truth (a contradiction is reported as contested, not resolved by fiat), and never synthesizes fewer than two sources (a lone source is summarized, not synthesized). | src/apothem/skills/source-synthesis/SKILL.md |
surgical-guard | Surgical-edit + reactive-guard skill — matched when the operator asks to 'edit surgically', 'make a minimal diff', 'guard this change', 'review the diff for quality', 'don't rewrite the whole file', or otherwise needs a precise, anchor-bounded mutation plus a post-edit quality pass before it lands. Stage 1 is a minimal-diff editing discipline: mutate through a managed-block or anchor-bounded edit (e.g. swap only the sentinel-delimited region, preserving the surrounding bytes and the authorship banner) — never a blunt whole-file overwrite where a scoped edit suffices. Stage 2 is a reactive, diff-based quality guard that reviews only the change's diff and loads only the rule references the diff's context matches (progressive disclosure), catching the systematic failure modes of generated changes: clean-code (swallowed errors, hardcoded success returns, hallucinated APIs, premature abstraction, silent contract changes), test (mock-boundary violations, duplicate bodies, hollow assertions, missing coverage of the changed behavior), docs (hallucinated symbols, broken samples, docs-vs-code drift). Not for: green-field file creation with no existing content to preserve (below the surgical threshold — routes to ordinary authoring); domain-framework linting with vendor-specific rule packs (out of scope — only the generalized clean-code/test/docs guard pattern is carried); producing the change's content via debugging or TDD (routes to dev-toolkit — this skill governs how the change lands and whether it passes). Harness-agnostic; deterministic output. Feeds the harness-directive surgical-manipulation mandate. | src/apothem/skills/surgical-guard/SKILL.md |
test-authoring | Behavior-first test authoring with strict Arrange-Act-Assert discipline — matched when the developer asks to 'write tests', 'add coverage', 'add a test for', 'test this function', 'cover this module', 'write a failing test', 'increase coverage', or any phrasing requesting new test cases that assert observable behavior against a unit under test. Discovers the unit's behavior contract (input domains, return shapes, side effects, raised exception types, edge boundaries — empty/null/zero/max/Unicode/off-by-one), writes failing AAA assertions before the implementation satisfies them, partitions coverage across the happy path, the edge boundaries, and the failure modes (with a regression test pinning any named bug), and reports residual coverage gaps for operator triage. Authors tests only — never edits, writes, or repairs the implementation under test, never executes the suite, never computes a coverage percentage, never edits coverage config, and never mocks the unit under test itself (mocks only at owned boundaries: adapters, injected ports). A request to fix a failing test by editing the implementation, scaffold a project, or run an existing suite is NOT a match. | src/apothem/skills/test-authoring/SKILL.md |
vuln-triage | Vulnerability and security-advisory triage — matched when the user references a 'CVE', cites a 'security advisory' (GHSA, OSV, vendor bulletin), asks to 'triage this vuln', or asks 'is this vulnerable'. Targets one named advisory per invocation: severity-classifies it with a CVSS-style band (critical 9.0–10.0 / high 7.0–8.9 / medium 4.0–6.9 / low 0.1–3.9 / none 0.0) plus the justifying vector string, maps it to the affected surface in the host tree against the resolved lockfile, determines reachability and exploitability (reachable / unreachable / conditional + applicable attack vector), routes remediation from the closed set {patch, upgrade, mitigate, accept} (or not-applicable when no affected surface exists), and records a seven-field triage record with cited rationale. The output drives the operator's remediation decision; the skill does NOT apply the remediation, construct or run exploit payloads, scan the repository, re-publish CVSS scores, or re-solve the dependency graph. The 'accept' route is a security ratification — when the host has not ratified an accepted-risk threshold, it blocks on inquiry. | src/apothem/skills/vuln-triage/SKILL.md |
workflow | General-purpose workflow-harnessing skill — matched when the operator states a mission via '/goal <<mission>>', asks to 'harness a workflow', 'run a multi-agent workflow', 'orchestrate agents', 'fan out and verify', 'critique and remediate', 'maximally elevate' a target, or otherwise hands off a non-trivial multi-step mission whose accomplishment benefits from genuinely-independent parallel work plus adversarial verification. Decomposes the mission, dispatches independent agents under named return contracts (non-overlapping scope, isolation where parallel writes collide, single-message parallel launch), subjects every load-bearing finding to an EXTREMELY-CRITIQUE refute-by-default verification pass (N independent critics per finding, default 3, distinct lenses — correctness/security/reproducibility/regression — survival only on non-refute majority) before it survives, is granted to identify and remediate defects beyond the literal mission (each disclosed per rules/disclosure-ledger.md), self-augments from current authoritative SOTA sources rather than memory alone, and emits a deterministic, byte-stable result. Multi-agent dispatch and continuous auto-execution are opt-in / confirmation-gated, never default-on — the canonical home for the '/goal <<mission>>' entry pattern. A single-step request that one direct tool call resolves is below this skill's threshold. | src/apothem/skills/workflow/SKILL.md |
नियम
व्यवहार-नियम सेट में समस्या-समाधान, कोड शिल्प, क्लीन आर्किटेक्चर, योजना, और गुणवत्ता को नियंत्रित करने वाले सदैव-सक्रिय और पथ-फ़िल्टर किए गए व्यवहार-नियम।
कमांड
प्लान-सूट पाइपलाइन को लागू करने वाली स्लैश-कमांड परिभाषाएँ, साथ ही कोड, आर्किटेक्चर, सुरक्षा, निर्भरताओं, और दस्तावेज़ीकरण समीक्षा के लिए ऑडिट।