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 TimSimpsonJr/magpie --skill pii-sweepgit clone --depth 1 https://github.com/TimSimpsonJr/magpieWrote 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/timsimpsonjr/magpie/pii-sweep)<a href="https://agentmods.dev/skills/timsimpsonjr/magpie/pii-sweep"><img src="https://agentmods.dev/badge/skills/timsimpsonjr/magpie/pii-sweep/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/skills/timsimpsonjr/magpie/pii-sweep"><img src="https://agentmods.dev/badge/skills/timsimpsonjr/magpie/pii-sweep.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 | $0.00136 | $0.02407 |
| Opus 5 | $0.00068 | $0.01203 |
| Sonnet 5 | $0.00027 | $0.00481 |
| Haiku 4.5 | $0.00014 | $0.00241 |
Grade A, and why
pii-sweep 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 10d 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 — 162 lines — stays where its author put it; the contents beside it link to each section on GitHub.
pii-sweep
Produce Magpie's authoritative PII-exposure tally over one FOIA / audit-log
free-text column (the reason / narrative / justification field): spaCy
en_core_web_lg PERSON NER plus structured-identifier regex, run over the
distinct values and weighted by row counts, splitting officials (named
for accountability) from PII that should have been sanitized. The aggregate tally
is the publishable headline; the matched texts stay LOCAL and feed redact-output
(Phase 7). This is the engine behind the recipe's pii check — that check is a
fast presence indicator; this is the authoritative count.
One deterministic engine does the work; the agent orchestrates the prep and the output:
scripts/pii_sweep.py::sweep(series, ...)— the distinct → classify → weight → tally pass (pure core; spaCy only at the lazy classifier edge).
The pure core is golden-tested with a fake classifier (no 400 MB model). The
verified spaCy facts, the NER-label scope, and the candidate pattern set live in
references/prior-art.md (the Phase 5 research gate) — consult it before changing
a model call or a regex. pii-sweep shares NO code with recipe.check_pii in
either direction (a drift test guards the overlapping patterns), so importing it
stays ML-free until sweep actually runs.
The pipeline
Run the steps in order.
-
Prep the source via
dataset-analyze(load_table → data_quality gate → derive). The sweep runs on ONE column: the cleaned free-text reason / narrative series. Gate on truncation FIRST — never sweep a silently-truncated export. -
Build the officials lexicon from the STRUCTURED column — not from the free text. Take the searcher / user / requesting-agency field (the structured identity already in the export) and pass its distinct values as
official_names=.sweepnormalizes each into name tokens and marks a PERSON spanofficialwhen an official's tokens are a subset of the span (robust to spaCy span over-extension) OR a rank/title token immediately precedes the span (Sgt,Officer,Deputy, ...). This is how an UNTITLED official ("Dana Wheeler ran the plate") is still attributed for accountability rather than counted as exposed PII.
What ships with it
1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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.
- 10d ago First seen · 162 lines · 136 tokens per session scan A 8517500edc80
pii-sweep is a skill published in the GitHub repository TimSimpsonJr/magpie (2 stars, last pushed 3mo ago), licensed MIT. It adds 136 tokens to every session and 2,407 once invoked, about $0.0007 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
pdf-table-extractor-brief
Produces a structured extraction plan and clean spreadsheet template for pulling tabular data out of a PDF document — identifying the table structure, defining column headers, flagging extraction pitfalls, and providing a ready-to-use template that ensures the data lands in a consistent, analysable format.
foia-request-writer
Drafts legally complete public records requests (federal FOIA and all 50 state laws), administrative appeals, and redaction challenge strategies for U.S. government records.
osint-tool-catalog
Produces a categorised catalog of open-source intelligence tools relevant to a journalist's investigation, with practical guidance on what each tool does, when to use it, and what its limitations are.
social-media-intelligence
Produces a structured open-source intelligence brief on a social media account or set of accounts, covering account authenticity analysis, narrative tracking, and coordination-detection patterns to support investigative reporting.
data-cleaning-brief
Writes clear, step-by-step instructions for cleaning a messy or inconsistent dataset — specifying exactly what needs to be standardised, corrected, or removed to make the data ready for analysis and publication.
foi-data-request
Drafts a Freedom of Information request specifically targeting datasets, databases, or digital records — written to maximise the chance of receiving complete, machine-readable data rather than summary PDFs or partial tables.