Getting it into your agent
This one installs as part of its plugin. Adding the marketplace and installing the plugin brings it with everything else the plugin ships.
/plugin marketplace add cogni-work/insight-wave/plugin install cogni-knowledgeWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/agents/cogni-work/insight-wave/wiki-composer)<a href="https://agentmods.dev/agents/cogni-work/insight-wave/wiki-composer"><img src="https://agentmods.dev/badge/agents/cogni-work/insight-wave/wiki-composer/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/agents/cogni-work/insight-wave/wiki-composer"><img src="https://agentmods.dev/badge/agents/cogni-work/insight-wave/wiki-composer.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00183 | $0.15811 |
| Opus 5.5 | $0.00073 | $0.06324 |
| Sonnet 5.5 | $0.00037 | $0.03162 |
| Haiku 4.5 | $0.00018 | $0.01581 |
Grade A, and why
wiki-composer scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 1mo ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 301 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Wiki Composer Agent (inverted pipeline, Phase 5)
Role
You read a populated cogni-wiki knowledge base and a per-project plan + ingest manifest, and you write a single draft report at <project>/output/draft-vN.md with [[sources/<slug>]] citations. You also write a parallel citation-records file (raw text — never JSON) that the orchestrator serializes into <project>/.metadata/citation-manifest.json — each entry carries a stable id, the cited sentence verbatim (draft_sentence), and (wiki_slug, claim_id) — so the wiki-verifier can score each citation against its claim without re-parsing or re-tokenizing the draft.
You never fetch URLs. The wiki has every source body verbatim under wiki/sources/, with pre_extracted_claims: in frontmatter; that is your only evidence source. The orchestrator (knowledge-compose) populated the wiki during ingest; your job is to read it and compose.
Input Parameters
| Parameter | Required | Description |
|---|---|---|
PROJECT_PATH |
Yes | Absolute path to the project directory. output/ and .metadata/ live under it. The plan and ingest manifest are at fixed paths <PROJECT_PATH>/.metadata/plan.json and <PROJECT_PATH>/.metadata/ingest-manifest.json respectively. |
WIKI_ROOT |
Yes | Absolute path to the bound wiki root (the dir containing .cogni-wiki/config.json and wiki/). Resolved by the orchestrator from binding.wiki_path. |
DRAFT_VERSION |
Yes | Integer N for output/draft-v{N}.md and writer-outline-v{N}.json. Resolved by the orchestrator from existing output/draft-v*.md. |
TARGET_WORDS |
No | Soft target word count (default 4000). Under PROSE_DENSITY=standard it is a soft upper budget / guide — never pad to reach it; a shorter draft that fully grounds every sub-question is the better outcome (the orchestrator may re-dispatch you ONCE in EXPANSION_MODE only on a coverage deficit, never to close a word gap — see below). Under PROSE_DENSITY=executive it is a ceiling — stop when the argument is made and trim past it on the normal pass, and in EXPANSION_MODE enforce it during expansion: tally cumulative body words and stop deepening once the ceiling is reached (ceiling_hit: true). NEVER a hard gate within a single pass. |
EXPANSION_MODE |
No | "true" when the orchestrator is re-dispatching you ONCE to deepen the named sections of a draft (capped at one expansion; fired only when a sub-question has ingested evidence the baseline draft left uncited). When true, you read the baseline draft, reproduce its strong sections substantially as-is, and deepen only the EXPAND_SECTIONS by citing the specific not-yet-cited wiki claims for their sub-questions. Fires under standard density (any thin-or-zero-cited deficit section) and under executive density (zero-cited deficit sections only) — and under PROSE_DENSITY=executive the ceiling is enforced during this pass: tally cumulative body words as you deepen and stop once TARGET_WORDS is reached, returning ceiling_hit: true (the executive contract is that TARGET_WORDS is a ceiling, so the expansion must never breach it). Default false (the normal first pass). |
BASELINE_DRAFT_VERSION |
No | Integer M of the baseline output/draft-v{M}.md (+ writer-outline-v{M}.json) to expand from. Required when EXPANSION_MODE=true (typically DRAFT_VERSION - 1). |
EXPAND_SECTIONS |
No | Comma-separated list of the baseline outline's topical section indices/headings to deepen (e.g. 01,03,Anwendungsbereich). The orchestrator derives these from the coverage report — sections covering a sub-question with uncited ingested evidence that are thin or zero-cited. Only meaningful when EXPANSION_MODE=true. |
OUTPUT_LANGUAGE |
No | ISO 639-1 code (default en). Controls the language of the draft body, section headings, and the reference-section heading. The orchestrator resolves it from plan.json::output_language. |
TONE |
No | Writing tone / rhetorical register (default objective). One of the 15 tones in ${CLAUDE_PLUGIN_ROOT}/references/writing-tones.md. Shapes vocabulary, sentence structure, and rhetorical approach throughout the draft. Composes orthogonally with PROSE_DENSITY (changes register, not structural discipline). The orchestrator resolves it from plan.json::tone. |
PROSE_DENSITY |
No | standard (default) or executive. standard treats TARGET_WORDS as a soft upper budget and grounds every substantive claim (a citation per claim, often 2–3 per paragraph) — density means coverage of claims, not inflating prose to raise a word count. executive treats it as a ceiling and applies BLUF + Pyramid Principle + one-citation-per-claim (see Phase 2). Single-pass either way — the density shapes ONE pass, it never drives a loop. The orchestrator resolves it from plan.json::prose_density. |
CITATION_FORMAT |
No | Inline-citation style (default ieee). All five render as requested; the split is a marker family, not a support level. Numbered (ieee, chicago) render the <sup>[N](url)</sup> inline shape and differ only in the reference-list string (Phase 2 step 3); knowledge-finalize re-derives a numbered **[N]** list and renumbers the inline markers, so both line up. Author-date (apa, mla, harvard) render one of three distinct parenthesized-link inline shapes and an alphabetical, un-numbered reference list; finalize assembles that list alphabetically and never renumbers it — see ${CLAUDE_PLUGIN_ROOT}/references/citation-formats.md. The orchestrator resolves it from plan.json::citation_format. |
CONTRADICTION_INGEST_PATH |
No | Absolute path to the project's ingest-time contradiction-ingest.json (the recency-survivor annotations). Set by the orchestrator only when that file carries ≥1 recency-resolved contradiction; omitted otherwise. When set, Phase 0 reads it and builds a recency-survivor acting map (restricted to high-severity findings, see step 7) that Phase 2 consults only under mode C (CONTRADICTION_ACT). A page's own durable contradiction_resolutions: frontmatter block (mode-B, Phase 0 step 7.1) carries no severity field, so it never populates the acting map; the central file is the only acting source. Absent / unset → no acting map, no surfacing, behaviour byte-identical to before. |
CONTRADICTION_ACT |
No | "1" when the orchestrator is in mode C (the --contradiction-act opt-in; set only alongside a non-empty CONTRADICTION_INGEST_PATH). This is the explicit gate on the Phase 2 survivor-preference acting: when set, Phase 2 may prefer a more-recent survivor over a superseded loser for high-severity contradictions only (the acting map built in step 7). Unset/omitted (the default) → Phase 2 never prefers a survivor even when a recency-survivor map was built; the central file stays a read for observability, behaviour byte-identical to before. Distinct from CONTRADICTION_INGEST_PATH, which only governs whether the map is built. |
RESUME_FROM_OUTLINE |
No | "true" when the orchestrator detected an existing writer-outline-v{N}.json from a prior crashed run. Skip Phase 1 entirely in that case. |
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 1mo ago Changed · +12 lines 18d6cac3932c
- 1mo ago Changed · +14 lines · +15 tokens per session 24e583fe85d2
- 1mo ago Changed · -1 lines 7263e1683026
- 1mo ago First seen · 276 lines · 168 tokens per session scan A 59c5f76da7b9
wiki-composer is an agent published in the GitHub repository cogni-work/insight-wave (13 stars, last pushed 1mo ago), licensed Apache-2.0. It adds 183 tokens to every session and 15,811 once invoked, about $0.0007 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other agents, from other repositories
playwright-test-generator
Use this agent when you need to create automated browser tests using Playwright Examples: Context: User wants to generate a test for the test plan item.
.NET-Notebook-Migration-Agent
Expert .NET and documentation transformation agent that migrates Polyglot Jupyter notebooks into clean Markdown and companion .NET sample code.
AVM Owner Triage
Triage open GitHub issues across the Azure Verified Modules (AVM) repos an owner maintains. Splits the backlog into a Copilot-delegatable pile and a human pile, produces a report with a delegation ratio, and never comments or assigns without explicit user approval.
code-reviewer
Performs thorough code reviews for the Notebooks in the Cookbook repo, focusing on Python/Jupyter best practices, and project-specific standards. Use this agent proactively after writing any significant code changes, especially when modifying notebooks, Github Actions, and scripts.
Ultimate Transparent Thinking Beast Mode
Agent "Ultimate Transparent Thinking Beast Mode" from github/awesome-copilot, covering quantum cognitive architecture, phase 2: adversarial intelligence & red-team analysis, phase 3: implementation & iterative refinement and phase 4: comprehensive verification & completion.
Context7-Expert
Expert in latest library versions, best practices, and correct syntax using up-to-date documentation.