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/adamayoung/tmdb/capture-knowledgenpx skills add adamayoung/TMDb --skill capture-knowledgegit clone --depth 1 https://github.com/adamayoung/TMDbWrote 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/adamayoung/tmdb/capture-knowledge)<a href="https://agentmods.dev/skills/adamayoung/tmdb/capture-knowledge"><img src="https://agentmods.dev/badge/skills/adamayoung/tmdb/capture-knowledge.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 5.1 | $0.00097 | $0.02371 |
| Opus 5 | $0.00048 | $0.01185 |
| Sonnet 5 | $0.00019 | $0.00474 |
| Haiku 4.5 | $0.00010 | $0.00237 |
Grade B, and why
capture-knowledge scanned grade B with 1 finding 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 6d 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.
Reads agent configuration directoriesmediumAgent snooping
.claude/, .codex/, .gemini/ hold keys, settings and other credentials a mod has no legitimate need for.
`knowledge/` entry heading, also grep `.claude/` and `CLAUDE.md` for How it starts
The opening of the file, as written. The whole thing — 158 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Capture Knowledge
Fold what you just learned into the committed knowledge base at knowledge/, so a
future session (or contributor) doesn't have to re-learn or re-discover it. Run
this before a PR — in /deliver it runs automatically pre-PR, so the
notes land in the same PR as the change.
What to capture (be selective)
Record only durable, project-specific, non-obvious facts — the test is "would a future me waste time without this?":
- Gotchas / quirks →
knowledge/gotchas.md— a trap you hit, a tooling surprise, anything that needed a web search or doc lookup to resolve. - Live-API behaviours →
knowledge/tmdb-api-notes.md— a nullable/absent field, an undocumented response shape, an enum value the docs omit. - Design decisions → a new ADR in
knowledge/decisions/(next number — take it fromdecisions/README.md's index, not a directory listing), usingdecisions/0000-template.md. Any non-obvious choice and its rationale — why this approach over the alternatives. Add the new row to that index. - Deferred breaking changes →
knowledge/next-major.md— any fix rejected or deferred because it is breaking. The skill-improvement log remembers the "no"; this file is what makes the deferred "yes" resurface at the next major. Reconcile its status line whenever you add an entry: it states which major window is open, and an entry filed against a window that already shipped is invisible at exactly the moment it should fire. A window is open until the version is tagged — an unreleasedCHANGELOGsection is not a release (git tag --list | sort -V | tail -1).
What NOT to capture
Mirror the discipline of a good memory — don't record:
- Anything already in the code,
CLAUDE.md, the topic docs under.claude/docs/, the DocC docs, or git history. - Facts that only mattered to this one task and won't recur.
- Restatements of the obvious. If asked to record something obvious, capture the non-obvious part (what surprised you) or skip it.
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.
- 6d ago First seen · 158 lines · 97 tokens per session scan B fa75bc1c4f38
capture-knowledge is a skill published in the GitHub repository adamayoung/TMDb (177 stars, last pushed 5d ago), licensed Apache-2.0. It adds 97 tokens to every session and 2,371 once invoked, about $0.0005 per session on Opus 5. A static security scan graded it B with 1 finding (reads agent configuration directories). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
mobile-compaction
Context compaction strategies for large Android codebases. Optimize token usage while preserving critical context.
mobile-memory
Memory persistence for mobile development context across sessions. Maintains project structure, dependencies, architecture, and test state.
mobile-instinct-v1
V1 instinct-based pattern capture for mobile development. Captures patterns as code is written with immediate feedback.
mobile-instinct-v2
V2 instinct-based observational learning. Analyzes sessions to extract reusable mobile development patterns across time.
continuous-learning-v2
Instinct-based learning system with confidence scoring for mobile patterns. Automatically extracts and evolves patterns.
Swift Patterns
Use this skill when working on Swift projects (SwiftPM packages, iOS/macOS apps) and you want consistent patterns for concurrency, structure, and safety.