本文へ移動
cccskills
無料GitHub で公開

oma-market

Research customer pain points, trends, and competitors through the OMA market engine. Use for market discovery or voice-of-customer analysis.

インストール方法を見る

含まれるファイル(9)

  • SKILL.md9.2 KB
  • resources/checklist.md876 B
  • resources/error-playbook.md1.4 KB
  • resources/execution-protocol.md2.6 KB
  • resources/frameworks/pestel.md2.8 KB
  • resources/frameworks/porters-5f.md3.0 KB
  • resources/frameworks/swot.md2.4 KB
  • resources/intent-rules.md2.5 KB
  • resources/output-laws.md2.3 KB

SKILL.md(原文)

インストールする前に、エージェントに与えられる指示の中身を確認できます。

Market Research Agent - Community Signal Intelligence

Scheduling

Goal

Run the upstream last30days research engine (always the latest release, managed by oma) for community-signal research, then frame the result for the user's intent (pain / trend / competitor / discovery) with strategic frameworks and save one brief under .agents/results/market/.

Intent signature

  • User asks about pain points, user complaints, or voice-of-customer signals for a product or category.
  • User asks what is trending, growing, or declining in a space this week or month.
  • User asks how one product compares to another in community sentiment or positioning.
  • User asks for discovery or exploratory market research on a topic, a person, a company, or a ticker.

When to use

  • Extracting real user pain points from community posts (Reddit with real upvotes and top comments, HN, X, Bluesky, GitHub Issues)
  • Detecting trends in a category over a window (--days 7|30|90|180)
  • Competitor sentiment analysis and SWOT / Porter's 5F positioning
  • Open-ended discovery research (--discover), person mode, hiring signals (--hiring-signals), follow-up drills (--drill)

When NOT to use

  • General web research without market framing -> use oma-search directly
  • Academic literature -> use oma-scholar
  • Live dashboards or scheduled monitoring -> oma schedule <action> wrapping this skill

Expected inputs

  • Topic string; optional --intent pain|trend|competitor|discovery (else classified per resources/intent-rules.md)
  • Optional window (--days), --vs <entity> (competitor), --frameworks auto|none|swot,5f,pestel
  • Any native last30days flag (see oma market run --help) — passed through verbatim

Expected outputs

  • Single markdown brief at .agents/results/market/{topic-slug}-{YYYYMMDD}.md
  • First line: the engine's badge (🌐 last30days v{VERSION} · synced {date}); body per the upstream OUTPUT CONTRACT; framework sections appended per intent; engine footer preserved
  • Raw engine artifacts under market.save_dir (default .agents/results/market/raw/)
outputs:
  - name: market-brief
    description: Single LAW-compliant markdown brief with framework sections
    artifact: ".agents/results/market/*.md"
    required: true

Dependencies

  • oma market resolve / oma market run — engine location, Python 3.12+ resolution, --save-dir default
  • The upstream SKILL.md at the resolved engine root (skillMd in oma market resolve --json) — the authoritative research contract
  • resources/intent-rules.md, resources/frameworks/, resources/output-laws.md, resources/execution-protocol.md

Control-flow features

  • oma market detect-trap gate before anything else (exit 2 = REFUSE, exit 4 = invalid)
  • Engine is always the latest release: oma market resolve refreshes the managed copy (throttled) and falls back to the cached copy offline; a pinned market.path / LAST30DAYS_HOME opts out
  • Sources needing keys/cookies auto-skip inside the engine; keyless sources (Reddit, HN, GitHub, Polymarket, arXiv, Techmeme, Digg, web) always run
  • Framework auto-toggle by intent (pain/trend → SWOT; competitor → SWOT + Porter's 5F; discovery → SWOT + PESTEL)

Structural Flow

Entry

  1. Run oma market detect-trap "<topic>". Exit 2 → surface the REFUSE reason and reframe suggestion, stop.
  2. Run oma market resolve --json. ok: false → report reason (missing engine → oma market update; missing Python → the install hint) and stop. Never fall back to WebSearch-only synthesis and present it as market research.
  3. Read the upstream contract at engine.skillMd top to bottom. It is long by design; do not skim. Treat engine.root as its SKILL_DIR.
  4. Classify intent per resources/intent-rules.md; map to engine flags and framework set.

Scenes

  1. PREPARE: detect-trap, resolve, read upstream SKILL.md, classify intent.
  2. UPSTREAM STEPS: follow the upstream SKILL.md exactly — Step 0 (first-run setup wizard, consent-driven), intent parsing, Step 0.45 (its own query-quality preflight), Step 0.5 / 0.55 (handle, subreddit, hashtag resolution when WebSearch is available), Step 0.75 (query plan). Skip only its "Runtime Preflight" Python-hunt block: oma market run already resolved the interpreter.
  3. RUN: wherever the upstream contract says "${LAST30DAYS_PYTHON}" "${SKILL_DIR}/scripts/last30days.py" <args>, run oma market run <args> with the same arguments (foreground, 5-minute timeout, --emit=compact). --save-dir is added automatically from market.save_dir unless you pass one.
  4. SYNTHESIZE: produce the brief exactly as the upstream OUTPUT CONTRACT dictates (badge first line, Ranked Evidence Clusters, LAWs). Then append the framework sections selected for the intent, using only clusters present in the engine output as evidence (resources/frameworks/).
  5. FINALIZE: run the self-check in resources/output-laws.md, write .agents/results/market/{topic-slug}-{YYYYMMDD}.md, preview the first 50 lines.

Transitions

  • --vs <entity> or "A vs B" phrasing → competitor intent → upstream COMPARISON flow (two passes + head-to-head as its contract specifies) → SWOT + Porter's 5F.
  • Person / company / ticker topics → upstream person / hiring-signals / StockTwits handling applies unchanged.
  • engine.status: stale → include the note in the report (research ran on the cached engine version).

Failure and recovery

  • detect-trap exit 2 → REFUSE; do not run the engine; --force only on explicit user reconfirmation.
  • oma market resolve not ok → stop with the reason; no engine run.
  • Engine non-zero exit → report stderr verbatim; do not synthesize from partial stdout unless the upstream contract says the emitted compact output is still valid.
  • Upstream Python-version gate / setup wizard messages → relay to the user exactly as the upstream contract instructs.

Exit

  • Success: brief written with badge, clusters, frameworks, and engine footer; path reported.
  • Partial: engine ran with skipped sources (footer lists them) — say so; never pad with invented evidence.

Logical Operations

Actions

ActionSSL primitiveEvidence
detect-trap preflightVALIDATETopic arg, trap pattern rules
Resolve engine + PythonCALL_TOOLoma market resolve --json
Read upstream contractREADengine.skillMd
Classify intentSELECTresources/intent-rules.md
Upstream pre-research stepsINFERUpstream SKILL.md Steps 0–0.75
Run engineCALL_TOOLoma market run <args>
Synthesize + frameworksWRITEUpstream OUTPUT CONTRACT, resources/frameworks/
Self-check + write briefWRITEresources/output-laws.md, .agents/results/market/

Tools and instruments

  • oma market detect-trap <topic> (preflight gate)
  • oma market resolve [--refresh|--offline] [--json] (engine + Python resolution; managed latest)
  • oma market update (force-refresh the managed engine)
  • oma market run <engine args…> (passthrough to the resolved upstream engine’s Python entry point)

Canonical command path

TOPIC="VS Code pain points"
oma market detect-trap "$TOPIC"
oma market resolve --json            # read .engine.skillMd, then follow it
# … upstream Steps 0 / 0.45 / 0.5 / 0.55 / 0.75 …
oma market run "$TOPIC" --plan "$QUERY_PLAN_FILE" --subreddits=vscode --emit=compact --save-suffix=v3

Resource scope

ScopeResource target
NETWORKInside the engine only (its per-source fetchers); GitHub for the managed engine refresh
LOCAL_FS~/.cache/oma-market/last30days/<tag>/ (engine), ~/.config/last30days/ (engine config, keys), .agents/results/market/ (brief + raw)
PROCESSoma market subcommands → the resolved upstream Python entry point

Preconditions

  • Topic passes detect-trap.
  • oma market resolve is ok (engine present; Python ≥ 3.12 found on PATH, via uv, or pinned with market.python / LAST30DAYS_PYTHON).

Effects and side effects

  • Writes the brief to .agents/results/market/{topic-slug}-{YYYYMMDD}.md and raw engine files to market.save_dir.
  • First run: the upstream setup wizard may write ~/.config/last30days/.env with user consent.
  • If Python 3.12+ is absent, oma market resolve stops before the upstream setup wizard. Relay its install hint; after an authorized interpreter installation, retry resolution. Do not expect the skipped upstream Python preflight to install it.

References

  • Execution protocol: resources/execution-protocol.md
  • Intent routing: resources/intent-rules.md
  • Output contract: resources/output-laws.md
  • Applicable framework: resources/frameworks/swot.md, resources/frameworks/porters-5f.md, or resources/frameworks/pestel.md (load only the selected framework)
  • Validation: resources/checklist.md
  • Recovery: resources/error-playbook.md
  • Upstream engine instructions: read the resolved engine.skillMd path from oma market resolve --json; upstream scripts are managed engine files, not bundled skill resources.

レビュー

まだレビューはありません。使ってみた感想をお寄せください。

同じリポジトリのスキル

概要と使いどころ

A fixture whose description uses a YAML folded scalar, which parses with an embedded trailing newline unless normalized before use in Markdown tables.

日本語の概要は準備中です。原文の説明を表示しています。

first-fluke/oh-my-agent1,3412026年10月12日 更新

Draft and revise academic prose against a rubric, evidence, and citation requirements. Use for essays, reports, literature reviews, or academic style audits.

日本語の概要は準備中です。原文の説明を表示しています。

first-fluke/oh-my-agent1,3412026年10月12日 更新

Evaluate system boundaries and architectural tradeoffs. Use for architecture decisions, design reviews, and ADRs.

日本語の概要は準備中です。原文の説明を表示しています。

first-fluke/oh-my-agent1,3412026年10月12日 更新

Architecture specialist for software/system design, module and service boundaries, tradeoff analysis, and stakeholder synthesis. Uses context-aware methods such as diagnostic routing, design-twice comparison, ATAM-style risk analysis, CBAM-style prioritization, and ADR-style decision records.

日本語の概要は準備中です。原文の説明を表示しています。

first-fluke/oh-my-agent1,3412026年10月12日 更新

Backend specialist for APIs, databases, authentication with clean architecture (Repository/Service/Router pattern). Use for API, endpoint, REST, database, server, migration, and auth work.

日本語の概要は準備中です。原文の説明を表示しています。

first-fluke/oh-my-agent1,3412026年10月12日 更新

Implement server APIs, authentication, and application data access. Schema modeling and query tuning use oma-db.

日本語の概要は準備中です。原文の説明を表示しています。

first-fluke/oh-my-agent1,3412026年10月12日 更新

first-fluke のスキルをすべて見る

このスキルの問題を報告する