feat: execpolicy v2 (#6467)
## Summary
- Introduces the `codex-execpolicy2` crate.
- This PR covers only the prefix-rule subset of the planned execpolicy
v2 language; a richer language will follow.
## Policy
- Policy language centers on `prefix_rule(pattern=[...], decision?,
match?, not_match?)`, where `pattern` is an ordered list of tokens; any
element may be a list to denote alternatives. `decision` defaults to
`allow`; valid values are `allow`, `prompt`, and `forbidden`. `match` /
`not_match` hold example commands that are tokenized and validated at
load time (think of these as unit tests).
## Policy shapes
- Prefix rules use Starlark syntax:
```starlark
prefix_rule(
pattern = ["cmd", ["alt1", "alt2"]], # ordered tokens; list entries denote alternatives
decision = "prompt", # allow | prompt | forbidden; defaults to allow
match = [["cmd", "alt1"]], # examples that must match this rule (enforced at compile time)
not_match = [["cmd", "oops"]], # examples that must not match this rule (enforced at compile time)
)
```
## Response shapes
- Match:
```json
{
"match": {
"decision": "allow|prompt|forbidden",
"matchedRules": [
{
"prefixRuleMatch": {
"matchedPrefix": ["<token>", "..."],
"decision": "allow|prompt|forbidden"
}
}
]
}
}
```
- No match:
```json
"noMatch"
```
- `matchedRules` lists every rule whose prefix matched the command;
`matchedPrefix` is the exact prefix that matched.
- The effective `decision` is the strictest severity across all matches
(`forbidden` > `prompt` > `allow`).
---------
Co-authored-by: Michael Bolin <mbolin@openai.com>
2025-11-17 10:15:45 -08:00
|
|
|
use crate::decision::Decision;
|
|
|
|
|
use crate::rule::RuleMatch;
|
|
|
|
|
use crate::rule::RuleRef;
|
|
|
|
|
use multimap::MultiMap;
|
|
|
|
|
use serde::Deserialize;
|
|
|
|
|
use serde::Serialize;
|
|
|
|
|
|
|
|
|
|
#[derive(Clone, Debug)]
|
|
|
|
|
pub struct Policy {
|
|
|
|
|
rules_by_program: MultiMap<String, RuleRef>,
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
impl Policy {
|
|
|
|
|
pub fn new(rules_by_program: MultiMap<String, RuleRef>) -> Self {
|
|
|
|
|
Self { rules_by_program }
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
pub fn rules(&self) -> &MultiMap<String, RuleRef> {
|
|
|
|
|
&self.rules_by_program
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
pub fn check(&self, cmd: &[String]) -> Evaluation {
|
|
|
|
|
let rules = match cmd.first() {
|
|
|
|
|
Some(first) => match self.rules_by_program.get_vec(first) {
|
|
|
|
|
Some(rules) => rules,
|
|
|
|
|
None => return Evaluation::NoMatch,
|
|
|
|
|
},
|
|
|
|
|
None => return Evaluation::NoMatch,
|
|
|
|
|
};
|
|
|
|
|
|
|
|
|
|
let matched_rules: Vec<RuleMatch> =
|
|
|
|
|
rules.iter().filter_map(|rule| rule.matches(cmd)).collect();
|
|
|
|
|
match matched_rules.iter().map(RuleMatch::decision).max() {
|
|
|
|
|
Some(decision) => Evaluation::Match {
|
|
|
|
|
decision,
|
|
|
|
|
matched_rules,
|
|
|
|
|
},
|
|
|
|
|
None => Evaluation::NoMatch,
|
|
|
|
|
}
|
|
|
|
|
}
|
2025-11-17 16:44:41 -08:00
|
|
|
|
|
|
|
|
pub fn check_multiple<Commands>(&self, commands: Commands) -> Evaluation
|
|
|
|
|
where
|
|
|
|
|
Commands: IntoIterator,
|
|
|
|
|
Commands::Item: AsRef<[String]>,
|
|
|
|
|
{
|
|
|
|
|
let matched_rules: Vec<RuleMatch> = commands
|
|
|
|
|
.into_iter()
|
|
|
|
|
.flat_map(|command| match self.check(command.as_ref()) {
|
|
|
|
|
Evaluation::Match { matched_rules, .. } => matched_rules,
|
|
|
|
|
Evaluation::NoMatch => Vec::new(),
|
|
|
|
|
})
|
|
|
|
|
.collect();
|
|
|
|
|
|
|
|
|
|
match matched_rules.iter().map(RuleMatch::decision).max() {
|
|
|
|
|
Some(decision) => Evaluation::Match {
|
|
|
|
|
decision,
|
|
|
|
|
matched_rules,
|
|
|
|
|
},
|
|
|
|
|
None => Evaluation::NoMatch,
|
|
|
|
|
}
|
|
|
|
|
}
|
feat: execpolicy v2 (#6467)
## Summary
- Introduces the `codex-execpolicy2` crate.
- This PR covers only the prefix-rule subset of the planned execpolicy
v2 language; a richer language will follow.
## Policy
- Policy language centers on `prefix_rule(pattern=[...], decision?,
match?, not_match?)`, where `pattern` is an ordered list of tokens; any
element may be a list to denote alternatives. `decision` defaults to
`allow`; valid values are `allow`, `prompt`, and `forbidden`. `match` /
`not_match` hold example commands that are tokenized and validated at
load time (think of these as unit tests).
## Policy shapes
- Prefix rules use Starlark syntax:
```starlark
prefix_rule(
pattern = ["cmd", ["alt1", "alt2"]], # ordered tokens; list entries denote alternatives
decision = "prompt", # allow | prompt | forbidden; defaults to allow
match = [["cmd", "alt1"]], # examples that must match this rule (enforced at compile time)
not_match = [["cmd", "oops"]], # examples that must not match this rule (enforced at compile time)
)
```
## Response shapes
- Match:
```json
{
"match": {
"decision": "allow|prompt|forbidden",
"matchedRules": [
{
"prefixRuleMatch": {
"matchedPrefix": ["<token>", "..."],
"decision": "allow|prompt|forbidden"
}
}
]
}
}
```
- No match:
```json
"noMatch"
```
- `matchedRules` lists every rule whose prefix matched the command;
`matchedPrefix` is the exact prefix that matched.
- The effective `decision` is the strictest severity across all matches
(`forbidden` > `prompt` > `allow`).
---------
Co-authored-by: Michael Bolin <mbolin@openai.com>
2025-11-17 10:15:45 -08:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
#[derive(Clone, Debug, Eq, PartialEq, Serialize, Deserialize)]
|
|
|
|
|
#[serde(rename_all = "camelCase")]
|
|
|
|
|
pub enum Evaluation {
|
|
|
|
|
NoMatch,
|
|
|
|
|
Match {
|
|
|
|
|
decision: Decision,
|
|
|
|
|
#[serde(rename = "matchedRules")]
|
|
|
|
|
matched_rules: Vec<RuleMatch>,
|
|
|
|
|
},
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
impl Evaluation {
|
|
|
|
|
pub fn is_match(&self) -> bool {
|
|
|
|
|
matches!(self, Self::Match { .. })
|
|
|
|
|
}
|
|
|
|
|
}
|