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 skills/mfrostbutter/skill-forge/workflow-capturenpx skills add Mfrostbutter/skill-forge --skill workflow-capturegit clone --depth 1 https://github.com/Mfrostbutter/skill-forgeWrote 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/skills/mfrostbutter/skill-forge/workflow-capture)<a href="https://agentmods.dev/skills/mfrostbutter/skill-forge/workflow-capture"><img src="https://agentmods.dev/badge/skills/mfrostbutter/skill-forge/workflow-capture.svg" alt="Measured on agentmods" 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 | $0.00112 | $0.00616 |
| Opus 5 | $0.00056 | $0.00308 |
| Sonnet 5 | $0.00022 | $0.00123 |
| Haiku 4.5 | $0.00011 | $0.00062 |
Grade A, and why
workflow-capture 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 5d 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 — 58 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Workflow Capture
Maintain an append-only journal that survives compaction and /clear, capturing the signal (decisions, sequencing, corrections, dead ends) that should become a skill. Do not log keystrokes; the transcript already does that at the wrong altitude. Capture why, not just what.
At session start
- Determine the active workflow slug (kebab-case, e.g.
figma-site-redesign). - Look for
_worklogs/YYYY-MM-DD-<slug>/WORKLOG.md.- If it exists, read it fully before doing anything. It is your memory.
- If not, create the directory and a new WORKLOG.md with a one-line header naming the workflow and date.
Tag schema
Every entry carries exactly one tag:
- DECISION: what was done and why -> a procedure step
- SEQUENCE: X must precede Y -> ordered steps or preconditions
- CORRECTION: user overrode you -> a hard rule
- DEADEND: tried X, failed because Y -> a "do not do this" warning
- CONVENTION: naming, structure, params, defaults -> reference material
- DISCOVERY: tool quirk, API limit, gotcha -> environment notes
- CHECKPOINT: state snapshot before clear or compaction -> resume anchor
Entry format
Append, never rewrite history.
## [YYYY-MM-DD HH:MM] TAG
One-line summary.
Optional result and fix or lesson.
Tags: comma, separated, keywords
When to append
Append an entry the moment one of these happens:
- The user corrects or overrides something (CORRECTION).
- An approach fails and you switch (DEADEND).
- A non-obvious ordering constraint appears (SEQUENCE).
- You set a reusable convention (CONVENTION).
- A tool behaves unexpectedly (DISCOVERY).
Keep entries short. The lesson matters, not the prose.
Checkpoint and resume
Before any compaction or /clear, write a CHECKPOINT entry: current state, open threads, next intended action. On resume in a fresh window, read the worklog and continue from the last CHECKPOINT plus accumulated tags.
Why this matters
The context window is volatile and mostly noise. A durable tagged journal is the only thing that lets a workflow survive a context reset and later become a skill that makes the next run cheaper and more correct.
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.
- 5d ago First seen · 58 lines · 112 tokens per session scan A 9c5b5ad354a4
workflow-capture is a skill published in the GitHub repository Mfrostbutter/skill-forge (4 stars, last pushed 1mo ago), licensed MIT. It adds 112 tokens to every session and 616 once invoked, about $0.0006 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 2026-08-31.
Other skills, from other repositories
kn-init
Use at the start of a new session to read project docs, understand context, and see current state.
kn-extract
Use when extracting reusable patterns, decisions, failures, or knowledge into documentation.
trellis-session-insight
Reach into past AI conversation history through the trellis mem CLI. Use whenever the user asks 'how did we solve X last time', 'have we discussed this before', 'what was the decision on X', 'remind me what we did in this task', '上次怎么解的', '之前讨论过吗', '想起一段对话', or when starting a brainstorm that overlaps prior work…
project-butler
Project memory workflow for init/upgrade, profile-aware setup, end session, normal/full close, file organization, document archiving, language switching, versioned update logs, rule review, status, wiki sync, and context recovery. Use for /project-butler, setup/初始化, foundation setup, profile setup, end session/收工…
vc-context-discovery
Discover and load all relevant context for the current task. Lists feature group nested files with full paths, loads process/context/ files by domain routing. Called at the start of every agent session.
vc-generate-context
Generate or update the project's authoritative repository context at process/context/all-context.md. Use when repo context is missing, stale, or contradicted by code.