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 skills add cosmix/loom --skill loom-before-aftergit clone --depth 1 https://github.com/cosmix/loomWrote 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/cosmix/loom/loom-before-after)<a href="https://agentmods.dev/skills/cosmix/loom/loom-before-after"><img src="https://agentmods.dev/badge/skills/cosmix/loom/loom-before-after.svg" alt="Measured on agentmods" height="20"></a>- NVIDIA SkillSpector warn
SkillSpector: 1 finding, up to medium
These are SkillSpector’s own severities. On a checked sample its high-severity flags on skills were ~96% false positives — a documented command, a public API, a “never do X” rule — so we show them as a caution to read, not a verdict. Why →
- medium Data Exfiltration · line 176 Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.Fix: Verify the destination URL is trusted and necessary. Remove or replace with documented APIs. Ensure no secrets, tokens, or PII are transmitted.
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 | $0.00015 | $0.02340 |
| Opus 5 | $0.00008 | $0.01170 |
| Sonnet 5 | $0.00003 | $0.00468 |
| Haiku 4.5 | $0.00002 | $0.00234 |
Grade B, and why
loom-before-after scanned grade B with 2 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 4d 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.
Sends data to an external URLmediumData exfiltration
A POST to an outside endpoint may be telemetry or may be exfiltration; either way the mod talks to somewhere, and you should know where.
1. **Testing too broadly.** `acceptance: ["cargo test"]` for "add register endpoint" proves nothing specific — assert the endpoint responds: `curl -sf -X POST .../api/register -d '{"email":"[email protected]"}' | jq -e '.user_id'` Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
⚠ **`before_stage` runs in a FRESH worktree before any agent work** — the reproducer must already exist in the base branch (a committed failing test, an existing endpoint), not one the stage is about to write. If the tes How it starts
The opening of the file, as written. The whole thing — 196 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Before/After Verification
Overview
A delta-proof proves a stage caused a change, not merely that the end state is valid. Capture what is true BEFORE implementation and what must be true AFTER; the pair distinguishes "my stage made it work" from "it already worked" (and "my fix resolved it" from "already fixed"). Without it, a green acceptance can pass on code that was already correct — proving nothing.
⚠️
truthsis GONE as a standalone field. Behavioral "after state" commands now go inacceptance(Simple string, or Extended object for output matching). The explicit, automated delta-proof mechanism is thebefore_stage/after_stagefields — unchanged, stillVec<TruthCheck>. A top-leveltruths:block is rejected as an unknown field.
The delta pattern
| Case | BEFORE (pre-condition) | AFTER (post-condition) |
|---|---|---|
| New feature | reproducer FAILS (feature absent) | reproducer SUCCEEDS |
| Bug fix (inverted!) | reproducer SUCCEEDS (bug present) | reproducer FAILS (bug gone) |
| Behavior change | old behavior observed | new behavior observed |
For a bug fix the direction inverts: the reproducer "passing" means the bug is still there. before_stage with exit_code: 1 on the reproducer, after_stage with exit_code: 0.
before_stage / after_stage — the automated delta-proof (how loom runs them)
Both are Vec<TruthCheck>; the plan author writes them. Exact lifecycle:
before_stageruns after worktree creation, BEFORE the session spawns, and only while the stage's workspace is pristine. On a failed check → stage goes Blocked, session is not spawned (you asserted a pre-condition that didn't hold). Infrastructure errors are advisory (warn + continue). 30-s timeout per check.- Skipped on re-spawns. If the stage branch already has commits beyond its base, or the worktree has changes, loom logs
Skipping before-stage checksand spawns anyway. A pre-condition asserting "the feature is absent" is expected to fail once an earlier attempt built it, so re-running it after orphan recovery /loom stage retry/ crash retry would block the stage on its own progress with no session able to finish the work. Write pre-conditions for a virgin workspace; do not rely on them re-running.
- Skipped on re-spawns. If the stage branch already has commits beyond its base, or the worktree has changes, loom logs
after_stageruns duringloom stage complete, AFTERacceptancepasses. On failure → stage stays Executing; the agent must fix and re-complete. 30-s timeout.
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.
- 4d ago Changed · -33 tokens per session 23bc3bdfb5b7
- 8d ago First seen · 196 lines · 48 tokens per session scan B a4fc92ac5bb9
loom-before-after is a skill published in the GitHub repository cosmix/loom (54 stars, last pushed yesterday), licensed MIT. It adds 15 tokens to every session and 2,340 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it B with 2 findings (sends data to an external url, makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
press-clip
Turn a live article URL into a press clip that looks like the real coverage — the publication's own logo, fonts, photos and layout kept intact, the ads and clutter removed, and (for a roundup) just the client's section. Renders to PDF. You inspect each site and tailor the removal; the bundled script carries no…
story-origin-check
Recover the first public timestamp and canonical major coverage for a newsjacking signal, then decide whether newer coverage is the same story, a different story, or a materially new development.
coverage-tracker
Run a Google Alerts-style keyword coverage tracker. Uses news-search for recent keyword queries, lets the LLM dedupe and classify real features versus junk, stores decisions in SQLite, and alerts only on new real coverage.
prompt-proximity-architecture
Turn an approved measurement charter, ICPs, and buyer jobs into a budget-aware prompt coverage blueprint across proximity bands, aided status, information acts, journey states, roles, locales, evidence grades, partitions, and measurement lanes. Use before prompt wording to define required, optional, and prohibited…
coverage-tracker-setup
Set up a lightweight Google Alerts-style coverage tracker for any number of keywords. Creates a tracker config with each keyword and what it actually means, then hands recurrence to the user's agent harness.
prompt-set-qa
Gate a prompt universe for schema and provenance completeness, target or campaign contamination, evidence entailment, naturalness, one-concept clarity, architecture consistency, aided status, answer leakage, and semantic duplicates. Use after realistic prompt generation and before human panel selection.