input-validation
/input-validation
Validate untrusted input at every trust boundary with allowlists, canonicalization, and hard limits so malformed data never reaches logic. Use when accepting data from requests, files, uploads, or third-party APIs.
- Author
- Amey-Thakur ↗
- Repository
- AI-SKILLS ↗
- Licence
- MIT
- Last repository change
- 05.09.2026
Assessment and security
- Files in this package
- 1
- Scripts
- 0
- HTML comments
- 0
- URLs in SKILL.md
- 0
No signals detected. Repository changed on 05.09.2026, without references to retired models or APIs.
MIT, open: use and modify while retaining author attribution.
What it does and when to use it
Validate untrusted input at every trust boundary with allowlists, canonicalization, and hard limits so malformed data never reaches logic. Use when accepting data from requests, files, uploads, or third-party APIs.
Original author description
Validate untrusted input at every trust boundary with allowlists, canonicalization, and hard limits so malformed data never reaches logic. Use when accepting data from requests, files, uploads, or third-party APIs.
How it works
- Method
- Checks
- Boundaries
SKILL.md preview
---
name: input-validation
description: Validate untrusted input at every trust boundary with allowlists, canonicalization, and hard limits so malformed data never reaches logic. Use when accepting data from requests, files, uploads, or third-party APIs.
---
# Input validation
Input validation is the wall between the outside world and code that trusts
its arguments. It fails when validation is scattered, negative (blocklists),
or run after the value has already been used. Do it once, at the boundary,
before anything downstream reads the value.
## Method
1. **Validate at the boundary, not deep in the stack.** Check input where it
enters: the request handler, the deserializer, the file parser. Once a
value passes the wall, downstream code may assume it is well-formed.
Validation buried three calls deep leaves the first two calls exposed.
2. **Prefer allowlists over blocklists.** Define what is legal (a set of
country codes, `^[a-z0-9_]{3,32}$` for a username, an enum of statuses)
and reject the rest. Blocklists lose: attackers find the encoding, the
Unicode homoglyph, or the case you did not enumerate.
3. **Canonicalize before you check.** Normalize first, validate second, or
an attacker slips past with an alternate encoding. Decode
percent-encoding, resolve `..` and symlinks in paths, apply Unicode
NFC, lowercase the host. Validating the raw form then using the decoded
form is the classic path-traversal bug.
4. **Cap length, count, and depth.** Set a maximum on every string, array,
upload, and nested-JSON depth. A 2 GB "username" or a 10,000-level nested
object is a denial of service you accepted by omission. Reject at the
limit with a clear error.
5. **Check type and range, then parse to a typed value.** Confirm the
integer is an integer within bounds, the date parses, the enum is a
member. Convert to the real type at the boundary so downstream code
handles `int` and `Date`, not `str`.
6. **Fail closed with a specific error.** On invalid input, reject and stop;
do not coerce, truncate silently, or "fix" it. Say which field failed and
why, but never echo the raw bad value into an error page unescaped.
## Checks
- Does each field have an explicit rule, or does some input reach logic only
because nothing rejected it?
- Do you canonicalize before validating everywhere a value is later decoded?
- Is there a length or size cap on every unbounded input, including request
bodies?
## Boundaries
Input validation shrinks attack surface; it does not replace output encoding
or parameterized queries. A validated string can still be dangerous in a SQL
statement or an HTML page, so defend those sinks separately (see
sql-injection-defense and xss-defense). Ownership checks (does this user own
this order?) belong with authorization, not here.
Versions
| Repository | Score | Licence |
|---|---|---|
| This page · Amey-Thakur/AI-SKILLS | 100 | MIT |
| drvoss/everything-copilot-cli | 100 | MIT |
| jeremymorgan/code-review-skills | 90 | CC0-1.0 |
Database snapshot: 13.11.2026 · package version b35c63e99fa7 (SHA-256 digest of the entire package)
Downloads in the last 30 days: 0 · ranking
Report a problem with this skill: harmful content, incorrect descriptions or an author’s removal request.
Similar skills
- 90security-boundaries
Use when untrusted input, credentials, permissions or tenant boundaries may permit unintended access. - 100harden
Harden code where untrusted input enters the system. Use when implementing auth, handling user input, storing sensitive data, integrating external APIs, adding file uploads or LLM features, or crossing any trust boundary. Don't use for secret scanning (/safe-repo) or dependency CVEs (/deps-audit). - 100security-hardening
Audits a change or a system for reachable vulnerabilities by tracing untrusted input to its sinks and checking authorization at every new path. Use when code touches authentication, secrets, cryptography, file paths, shell execution, deserialization, or any input crossing a… - 100input-validation-injection
Apply when reviewing or writing code that processes untrusted input, constructs queries or commands, or handles user-supplied data. Covers SQL, LDAP, OS command injection, prototype pollution, and general validation strategy. - 100security-verification
Check that every user-input surface is validated and sanitized — forms, textareas, query and URL params, external data. Use when asked to verify input validation, audit sanitization, or confirm untrusted input is handled safely. - 100security-verification
Check that every user-input surface is validated and sanitized — forms, textareas, query and URL params, external data. Use when asked to verify input validation, audit sanitization, or confirm untrusted input is handled safely.