execkit 0.7.2

Stateful, structured, safe shell sessions for AI agents on real infrastructure.
Documentation
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
// SPDX-License-Identifier: Apache-2.0
//! Output budgets: per-stream shaping of command output so huge logs do not
//! blow an agent's context window. Pure logic - the pipeline (grep -> line-keep
//! -> char-cap) operates on already ANSI-stripped, secret-redacted text and is
//! applied independently to stdout and stderr by `Session`.

use crate::error::{Error, Result};
use crate::output::bound;
use serde::{Deserialize, Serialize};

/// A per-stream output-shaping pipeline. `Budget::default()` is a no-op.
#[derive(Debug, Clone, Default, PartialEq, Eq, Serialize, Deserialize)]
pub struct Budget {
    /// 1. Keep only lines matching this regex (plus context). None = keep all.
    pub grep: Option<Grep>,
    /// 2. Bound the survivors to a line window.
    pub keep: Keep,
    /// 3. Final char backstop for THIS call. None falls back to the session cap.
    pub max_chars: Option<usize>,
}

/// A regex line filter with symmetric context (like `grep -C`).
#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize)]
pub struct Grep {
    pub pattern: String,
    /// Lines of context kept on EACH side of a match. 0 = match only.
    pub context: usize,
}

/// How to bound the lines that survive the grep filter.
#[derive(Debug, Clone, Copy, Default, PartialEq, Eq, Serialize, Deserialize)]
pub enum Keep {
    #[default]
    All,
    Tail(usize),
    Head(usize),
    HeadTail(usize, usize),
}

impl Budget {
    pub fn tail(n: usize) -> Self {
        Budget {
            keep: Keep::Tail(n),
            ..Default::default()
        }
    }
    pub fn head(n: usize) -> Self {
        Budget {
            keep: Keep::Head(n),
            ..Default::default()
        }
    }
    pub fn head_tail(head: usize, tail: usize) -> Self {
        Budget {
            keep: Keep::HeadTail(head, tail),
            ..Default::default()
        }
    }
    /// Grep filter (context 0, keep all matches). Chain `.context`/`.keep`/`.max_chars`.
    pub fn grep(pattern: impl Into<String>) -> Self {
        Budget {
            grep: Some(Grep {
                pattern: pattern.into(),
                context: 0,
            }),
            ..Default::default()
        }
    }
    /// Set grep context lines (no-op if there is no grep filter).
    pub fn context(mut self, n: usize) -> Self {
        if let Some(g) = self.grep.as_mut() {
            g.context = n;
        }
        self
    }
    pub fn keep(mut self, keep: Keep) -> Self {
        self.keep = keep;
        self
    }
    pub fn max_chars(mut self, n: usize) -> Self {
        self.max_chars = Some(n);
        self
    }
}

#[cfg(test)]
mod builder_tests {
    use super::*;

    #[test]
    fn constructors_build_expected_budgets() {
        assert_eq!(Budget::tail(10).keep, Keep::Tail(10));
        assert_eq!(Budget::head(5).keep, Keep::Head(5));
        assert_eq!(Budget::head_tail(2, 3).keep, Keep::HeadTail(2, 3));
        assert_eq!(Budget::default(), Budget::default());
        assert!(Budget::default().grep.is_none());
    }

    #[test]
    fn grep_builder_chains() {
        let b = Budget::grep("err")
            .context(2)
            .keep(Keep::Tail(50))
            .max_chars(100);
        assert_eq!(b.grep.as_ref().unwrap().pattern, "err");
        assert_eq!(b.grep.as_ref().unwrap().context, 2);
        assert_eq!(b.keep, Keep::Tail(50));
        assert_eq!(b.max_chars, Some(100));
    }

    #[test]
    fn context_without_grep_is_noop() {
        let b = Budget::tail(5).context(3);
        assert!(b.grep.is_none());
    }
}

/// Indices of `content` to keep: every line matching `re`, plus `context` lines
/// on each side, merged and sorted. Empty input -> empty.
fn grep_keep_indices(content: &[&str], re: &regex::Regex, context: usize) -> Vec<usize> {
    let mut keep = vec![false; content.len()];
    for (i, line) in content.iter().enumerate() {
        if re.is_match(line) {
            let lo = i.saturating_sub(context);
            let hi = (i + context).min(content.len().saturating_sub(1));
            for k in keep.iter_mut().take(hi + 1).skip(lo) {
                *k = true;
            }
        }
    }
    (0..content.len()).filter(|&i| keep[i]).collect()
}

#[cfg(test)]
mod grep_tests {
    use super::*;

    fn idx(text: &str, pat: &str, ctx: usize) -> Vec<usize> {
        let content: Vec<&str> = text.lines().collect();
        grep_keep_indices(&content, &regex::Regex::new(pat).unwrap(), ctx)
    }

    #[test]
    fn match_only_no_context() {
        assert_eq!(idx("a\nERR\nb\nERR\nc", "ERR", 0), vec![1, 3]);
    }

    #[test]
    fn context_expands_and_merges_overlaps() {
        // matches at 1 and 3, context 1 -> {0,1,2} and {2,3,4} merge to 0..=4
        assert_eq!(idx("a\nERR\nb\nERR\nc", "ERR", 1), vec![0, 1, 2, 3, 4]);
    }

    #[test]
    fn context_clamps_at_edges() {
        assert_eq!(idx("ERR\nb\nc", "ERR", 5), vec![0, 1, 2]);
    }

    #[test]
    fn no_match_is_empty() {
        assert!(idx("a\nb\nc", "ZZZ", 2).is_empty());
    }
}

/// Reduce a sorted index list to the head/tail window requested by `keep`.
fn keep_subset(idx: &[usize], keep: Keep) -> Vec<usize> {
    let n = idx.len();
    match keep {
        Keep::All => idx.to_vec(),
        Keep::Tail(k) => idx[n.saturating_sub(k)..].to_vec(),
        Keep::Head(k) => idx[..k.min(n)].to_vec(),
        Keep::HeadTail(h, t) => {
            // `h`/`t` are untrusted (usize straight from JSON); use saturating
            // math and clamp the slice bounds so extreme values can't overflow or
            // index out of range - they just collapse to "keep everything".
            if h.saturating_add(t) >= n {
                idx.to_vec()
            } else {
                let mut v = idx[..h.min(n)].to_vec();
                v.extend_from_slice(&idx[n - t.min(n)..]);
                v
            }
        }
    }
}

/// Render the kept lines with `... N lines elided ...` markers for every gap
/// (leading, internal, trailing) relative to the original `total` lines.
fn render(content: &[&str], kept: &[usize], total: usize) -> String {
    if kept.is_empty() {
        return if total == 0 {
            String::new()
        } else {
            format!("... {total} lines elided ...")
        };
    }
    let mut out: Vec<String> = Vec::new();
    let marker = |n: usize| format!("... {n} lines elided ...");
    if kept[0] > 0 {
        out.push(marker(kept[0]));
    }
    for (pos, &i) in kept.iter().enumerate() {
        if pos > 0 {
            let prev = kept[pos - 1];
            if i > prev + 1 {
                out.push(marker(i - prev - 1));
            }
        }
        out.push(content[i].to_string());
    }
    let last = *kept.last().unwrap();
    if last < total - 1 {
        out.push(marker(total - 1 - last));
    }
    out.join("\n")
}

#[cfg(test)]
mod keep_tests {
    use super::*;

    #[test]
    fn tail_head_headtail_windows() {
        let idx: Vec<usize> = (0..10).collect();
        assert_eq!(keep_subset(&idx, Keep::Tail(3)), vec![7, 8, 9]);
        assert_eq!(keep_subset(&idx, Keep::Head(2)), vec![0, 1]);
        assert_eq!(keep_subset(&idx, Keep::HeadTail(2, 2)), vec![0, 1, 8, 9]);
        assert_eq!(keep_subset(&idx, Keep::All), idx);
    }

    #[test]
    fn keep_windows_clamp() {
        let idx: Vec<usize> = (0..3).collect();
        assert_eq!(keep_subset(&idx, Keep::Tail(99)), idx); // n>len -> all
        assert_eq!(keep_subset(&idx, Keep::Head(99)), idx);
        assert_eq!(keep_subset(&idx, Keep::HeadTail(2, 2)), idx); // h+t>=n -> all
        assert!(keep_subset(&idx, Keep::Tail(0)).is_empty());
    }

    #[test]
    fn keep_headtail_extreme_counts_do_not_panic() {
        // Untrusted usize from JSON: must not overflow `h+t` nor index OOB.
        let idx: Vec<usize> = (0..20).collect();
        assert_eq!(keep_subset(&idx, Keep::HeadTail(usize::MAX, 1)), idx);
        assert_eq!(keep_subset(&idx, Keep::HeadTail(1, usize::MAX)), idx);
        assert_eq!(
            keep_subset(&idx, Keep::HeadTail(usize::MAX, usize::MAX)),
            idx
        );
    }

    #[test]
    fn render_marks_leading_internal_trailing_gaps() {
        let content: Vec<&str> = "l0\nl1\nl2\nl3\nl4".lines().collect();
        // keep indices 1 and 3 of 5 total
        let r = render(&content, &[1, 3], 5);
        assert_eq!(
            r,
            "... 1 lines elided ...\nl1\n... 1 lines elided ...\nl3\n... 1 lines elided ..."
        );
    }

    #[test]
    fn render_empty_kept() {
        let content: Vec<&str> = "a\nb".lines().collect();
        assert_eq!(render(&content, &[], 2), "... 2 lines elided ...");
        assert_eq!(render(&[], &[], 0), "");
    }
}

/// What a budget did to one stream. Present on `ExecResult` only when a
/// non-default budget was applied.
#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize)]
pub struct StreamReport {
    /// Short label of the pipeline that ran: "all","tail","head","head_tail",
    /// "grep", or composed like "grep+tail".
    pub mode: String,
    pub lines_total: usize,
    pub lines_kept: usize,
}

/// Per-stream reports for a shaped `ExecResult`.
#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize)]
pub struct BudgetReport {
    pub stdout: StreamReport,
    pub stderr: StreamReport,
}

fn keep_label(keep: Keep) -> Option<&'static str> {
    match keep {
        Keep::All => None,
        Keep::Tail(_) => Some("tail"),
        Keep::Head(_) => Some("head"),
        Keep::HeadTail(_, _) => Some("head_tail"),
    }
}

/// Max grep pattern length. The pattern is agent/MCP-controlled (untrusted), so
/// bound it: a huge pattern can be expensive to COMPILE even though matching is
/// linear. Paired with a compiled-program size limit below.
const MAX_GREP_PATTERN_LEN: usize = 4096;

/// Compile a grep pattern defensively (length cap + bounded compiled size) so an
/// untrusted caller cannot submit one that is pathologically costly to compile.
pub(crate) fn compile_grep(pattern: &str) -> Result<regex::Regex> {
    if pattern.len() > MAX_GREP_PATTERN_LEN {
        return Err(Error::Budget(format!(
            "grep pattern too long ({} bytes; max {MAX_GREP_PATTERN_LEN})",
            pattern.len()
        )));
    }
    regex::RegexBuilder::new(pattern)
        .size_limit(1 << 20) // 1 MiB compiled-program cap
        .build()
        .map_err(|e| Error::Budget(format!("invalid grep pattern: {e}")))
}

/// Shape one already-redacted stream. Returns (text, report, char_capped).
pub(crate) fn apply(
    text: &str,
    budget: &Budget,
    fallback_max_chars: usize,
) -> Result<(String, StreamReport, bool)> {
    let content: Vec<&str> = text.lines().collect();
    let total = content.len();

    let mut idx: Vec<usize> = (0..total).collect();
    let mut parts: Vec<&str> = Vec::new();
    if let Some(g) = &budget.grep {
        let re = compile_grep(&g.pattern)?;
        idx = grep_keep_indices(&content, &re, g.context);
        parts.push("grep");
    }
    idx = keep_subset(&idx, budget.keep);
    if let Some(l) = keep_label(budget.keep) {
        parts.push(l);
    }
    let lines_kept = idx.len();
    let rendered = render(&content, &idx, total);

    let cap = budget.max_chars.unwrap_or(fallback_max_chars);
    let (capped_text, capped) = bound(&rendered, cap);

    let mode = if parts.is_empty() {
        "all".to_string()
    } else {
        parts.join("+")
    };
    Ok((
        capped_text,
        StreamReport {
            mode,
            lines_total: total,
            lines_kept,
        },
        capped,
    ))
}

#[cfg(test)]
mod apply_tests {
    use super::*;

    #[test]
    fn default_budget_is_passthrough_until_cap() {
        let (t, r, capped) = apply("a\nb\nc", &Budget::default(), 1000).unwrap();
        assert_eq!(t, "a\nb\nc");
        assert_eq!(r.mode, "all");
        assert_eq!((r.lines_total, r.lines_kept), (3, 3));
        assert!(!capped);
    }

    #[test]
    fn tail_keeps_last_n_with_report() {
        let text = (0..100)
            .map(|i| i.to_string())
            .collect::<Vec<_>>()
            .join("\n");
        let (t, r, _) = apply(&text, &Budget::tail(3), 100_000).unwrap();
        assert!(t.starts_with("... 97 lines elided ..."));
        assert!(t.ends_with("97\n98\n99"));
        assert_eq!(
            (r.lines_total, r.lines_kept, r.mode.as_str()),
            (100, 3, "tail")
        );
    }

    #[test]
    fn grep_then_tail_composes() {
        let mut lines: Vec<String> = (0..50).map(|i| format!("info {i}")).collect();
        lines[10] = "ERROR ten".into();
        lines[40] = "ERROR forty".into();
        let text = lines.join("\n");
        let b = Budget::grep("ERROR").keep(Keep::Tail(1));
        let (t, r, _) = apply(&text, &b, 100_000).unwrap();
        assert!(t.contains("ERROR forty"));
        assert!(!t.contains("ERROR ten")); // tail(1) of the 2 matches
        assert_eq!((r.lines_kept, r.mode.as_str()), (1, "grep+tail"));
    }

    #[test]
    fn invalid_regex_errors() {
        let err = apply("x", &Budget::grep("("), 100).unwrap_err();
        assert!(matches!(err, Error::Budget(_)));
    }

    #[test]
    fn oversized_pattern_rejected() {
        let huge = "a".repeat(MAX_GREP_PATTERN_LEN + 1);
        let err = apply("x", &Budget::grep(huge), 100).unwrap_err();
        match err {
            Error::Budget(m) => assert!(m.contains("too long"), "{m}"),
            other => panic!("expected Budget error, got {other:?}"),
        }
    }

    #[test]
    fn char_cap_flags_capped() {
        let text = "x".repeat(1000);
        let (_t, _r, capped) = apply(&text, &Budget::default(), 10).unwrap();
        assert!(capped);
    }
}