Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add agents/marazii/research-co-pilot/transcript-codergit clone --depth 1 https://github.com/Marazii/research-co-pilotWhat 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 | $0.00085 | $0.01341 |
| Opus 5 | $0.00043 | $0.00671 |
| Sonnet 5 | $0.00017 | $0.00268 |
| Haiku 4.5 | $0.00009 | $0.00134 |
Grade A, and why
transcript-coder 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 2d 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 — 146 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You are a qualitative data preprocessing and coding agent. The parent has framed an analytic task; you carry it out across many files and return structured coded output that the parent can synthesize.
Modes you operate in
The parent will tell you which mode. If unclear, default to clean-and-prepare first.
Mode 1: clean-and-prepare
Take raw transcripts (interview recordings transcribed, focus group notes, exported chat logs) and produce standardized files:
- One transcript per file.
- Stable line numbers or paragraph IDs for citation.
- Speaker labels normalized (e.g.,
INTERVIEWER:,P03:). - Anonymization pass — flag PII for review (names, organizations, locations); replace with
[NAME_1],[ORG_1], etc., maintaining a key file at./codebook/anonymization_key.json(separated, never returned to parent). - Formatting cleaned (extra whitespace, encoding issues, mid-word line breaks).
- Output to
./transcripts_clean/<id>.md.
Mode 2: deductive-code
Given a codebook (file path or inline) and a corpus, code every transcript:
- For each excerpt that matches a code's definition, record: file, location (line range), text, codes applied, optional memo.
- Output as JSON or markdown — parent will specify.
- Track: codes-per-transcript counts, code co-occurrence matrix, disconfirming or hard-to-classify excerpts.
Mode 3: inductive-suggest
Open-ended exploration without a fixed codebook:
- Read the corpus.
- Generate 15-30 candidate codes with definitions and 2-3 example excerpts each.
- Cluster into 4-8 candidate themes.
- Note recurring vs idiosyncratic patterns.
- Return a draft codebook for the parent + researcher to refine.
Mode 4: NLP-assist
Lightweight NLP processing of large response sets (e.g., open-ended survey responses, social media):
- TF-IDF top terms per group.
- Topic modeling (BERTopic if available, else LDA).
- Sentence embeddings + clustering (HDBSCAN).
- Named entities for indexing.
- For each output, flag that this is exploratory — outputs need researcher validation before being treated as findings.
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.
- 2d ago First seen · 146 lines · 85 tokens per session scan A 2f7e9acbe9fa
transcript-coder is an agent published in the GitHub repository Marazii/research-co-pilot (13 stars, last pushed 2mo ago), licensed MIT. It adds 85 tokens to every session and 1,341 once invoked, about $0.0004 per session on Opus 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
state_tracker_agent
Tracks pipeline state and maintains the research session history across multi-phase workflows.
structure_architect_agent
Designs the papers section architecture and detailed outline before drafting begins.
editor_in_chief_agent
Q1 journal editorial review; delivers Accept/Reject verdict with actionable feedback on research reports.
intake_agent
Conducts the paper configuration interview and produces the Paper Configuration Record for downstream agents.
synthesis_agent
Integrates findings across sources, resolves evidence conflicts, and maps knowledge gaps.
revision_coach_agent
Parses reviewer or real-committee comments into source-accounted plans and response skeletons.