release: agf v0.14.0 (#77)

* feat: harden session discovery and add Prime Agent

* fix: make Windows executable tests portable

* test: use Windows-safe temporary paths

* test: respect Windows timestamp resolution

Author: subinium

.github/workflows/ci.yml

@@ -12,6 +12,24 @@ concurrency:
   cancel-in-progress: true
 
 jobs:
+  msrv:
+    name: MSRV 1.88
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+      - uses: dtolnay/rust-toolchain@1.88.0
+      - uses: Swatinem/rust-cache@v2
+      - run: cargo check --all-targets --all-features --locked
+
+  audit:
+    name: Security audit
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+      - uses: rustsec/audit-check@v2.0.0
+        with:
+          token: ${{ secrets.GITHUB_TOKEN }}
+
   clippy:
     name: Clippy
     runs-on: ubuntu-latest
@@ -23,7 +41,7 @@ jobs:
       - uses: Swatinem/rust-cache@v2
       # --all-targets matches the local gate (lints tests too); clippy
       # subsumes cargo check, so there is no separate check job.
-      - run: cargo clippy --all-targets --locked -- -D warnings
+      - run: cargo clippy --all-targets --all-features --locked -- -D warnings
 
   fmt:
     name: Format
@@ -42,7 +60,7 @@ jobs:
       - uses: actions/checkout@v4
       - uses: dtolnay/rust-toolchain@stable
       - uses: Swatinem/rust-cache@v2
-      - run: cargo test --locked
+      - run: cargo test --all-features --locked
 
   windows:
     name: Windows
@@ -54,5 +72,5 @@ jobs:
           components: clippy
       - uses: Swatinem/rust-cache@v2
       # Lint here too so #[cfg(windows)]-gated code is covered.
-      - run: cargo clippy --all-targets --locked -- -D warnings
-      - run: cargo test --locked
+      - run: cargo clippy --all-targets --all-features --locked -- -D warnings
+      - run: cargo test --all-features --locked

.github/workflows/release.yml

@@ -39,6 +39,9 @@ jobs:
           - target: x86_64-unknown-linux-gnu
             os: ubuntu-latest
             artifact: agf-x86_64-unknown-linux-gnu
+          - target: aarch64-unknown-linux-gnu
+            os: ubuntu-24.04-arm
+            artifact: agf-aarch64-unknown-linux-gnu
           - target: x86_64-pc-windows-msvc
             os: windows-latest
             artifact: agf-x86_64-pc-windows-msvc
@@ -98,8 +101,8 @@ jobs:
           shopt -s nullglob
           files=(*/*.tar.gz */*.zip)
           # Keep in sync with the build matrix size.
-          if [ "${#files[@]}" -ne 4 ]; then
-            echo "expected 4 release archives, found ${#files[@]}: ${files[*]}" >&2
+          if [ "${#files[@]}" -ne 5 ]; then
+            echo "expected 5 release archives, found ${#files[@]}: ${files[*]}" >&2
             exit 1
           fi
           for f in "${files[@]}"; do

CHANGELOG.md

@@ -2,6 +2,39 @@
 
 ## [Unreleased]
 
+## [0.14.0] - 2026-08-15
+
+### Added
+
+- **Prime Agent support** — discovers flat Prime Agent v3 JSONL sessions, resumes exact IDs with `prime-agent --resume`, honors environment/global session-root settings, and sorts by durable user/assistant activity. Named sessions, git state, and current recap metadata are surfaced. Deletion is fail-closed because upstream exposes no public safe delete command for active daemon sessions.
+- **Kiro v3 support** — discovers metadata/event pairs under `$KIRO_HOME/sessions/cli` in addition to the v2 SQLite store. Kiro resume now uses the selected conversation ID via `--resume-id`.
+- **Codex non-interactive filtering** — subagent and exec threads are hidden by default and can be included with `--include-non-interactive` or the matching setting.
+- Linux ARM64 release artifacts, an MSRV 1.88 CI lane, and a RustSec audit gate.
+
+### Fixed
+
+- Codex scans are now strictly read-only. Partial rollout walks can no longer delete valid rows from Codex-owned SQLite databases.
+- Cache invalidation uses nanosecond/size/path fingerprints, follows deeper Codex stores, and watches SQLite WAL files plus every secondary scanner input. Stale rows remain visible when a refresh fails and failed scans are never cached as successful emptiness.
+- Project View stores stable `(agent, session ID)` identities and rebuilds groups after streamed results, preventing wrong-session actions and out-of-bounds panics. `max_sessions` is now display-only and never truncates persistent cache data.
+- Time, name, and agent sorts use deterministic total ordering; timestamps are normalized globally and legacy Codex activity uses rollout mtime.
+- Exact quick-resume behavior: summary queries work, agent filters push down to one scanner, invalid permission modes fail clearly, and current Codex approval/sandbox flags replace removed aliases.
+- Terminal behavior: non-TTY TUI/watch calls fail cleanly, the session-row mouse offset is correct, shell wrappers preserve agent exit status, cwd-independent Hermes resumes execute, `NO_COLOR`/`TERM=dumb` are honored, and untrusted display metadata cannot emit terminal controls.
+- Deletes use collision-safe atomic writes and exact session matching; Claude append-only history is never rewritten while an agent may be writing it, Gemini duplicates are removed across stores, and Hermes prefix siblings survive.
+- Stats windows are cumulative and future/invalid timestamps are separate. Watch opens from cache immediately, preserves selection by identity, rejects zero intervals, and avoids repeated missing-`pgrep` probes.
+
+### Performance
+
+- Cursor store databases are indexed once instead of scanned once per transcript.
+- Pi, Cursor, Kiro, and Prime Agent JSONL parsing has total and per-line allocation bounds; Kiro v2 previews no longer materialize unlimited conversation blobs.
+- CLI commands reuse fresh per-agent cache entries and scan only stale/requested agents.
+- Text truncation is grapheme-aware, preserving emoji/ZWJ clusters while maintaining terminal-column alignment.
+
+### Changed
+
+- SuperLightTUI 0.20.1 → 0.22.3, including upstream terminal lifecycle, suspend/resume, bidi, and Unicode fixes.
+- Rusqlite is pinned to 0.39.0 so the declared Rust 1.88 MSRV remains buildable; the 0.40 transitive build script currently requires Rust 1.95 without declaring it.
+- Cache schema 8 → 9. Pin and summary state now use composite agent/session identities while legacy bare-ID pins remain readable.
+
 ## [0.13.1] - 2026-07-31
 
 Post-release audit of 0.13.0. No new features and no CLI changes; everything below is a correctness, durability, or performance fix found by reading the 0.13.0 diff and then the rest of the tree.

Cargo.lock

@@ -1,30 +1,33 @@
 [package]
 name = "agf"
-version = "0.13.1"
+version = "0.14.0"
 edition = "2024"
 # 1.88 = let-chains (edition 2024); also covers usize::is_multiple_of (1.87).
 rust-version = "1.88"
-description = "Find and resume local AI coding-agent sessions across Claude Code, Codex, Gemini, Cursor CLI, OpenCode, Kiro, pi, Hermes, Oh My Pi, and Yolop"
+description = "Find and resume local AI coding-agent sessions across Claude Code, Codex, Prime Agent, Gemini, Cursor CLI, OpenCode, Kiro, pi, Hermes, Oh My Pi, and Yolop"
 license = "MIT"
 repository = "https://github.com/subinium/agf"
 keywords = ["tui", "cli", "agent", "session", "claude"]
 categories = ["command-line-utilities"]
 readme = "README.md"
 
 [dependencies]
-superlighttui = "0.20"
+superlighttui = "0.22.3"
 serde = { version = "1", features = ["derive"] }
 serde_json = "1"
-rusqlite = { version = "0.40", features = ["bundled"] }
+# 0.40's libsqlite3-sys currently uses std::cfg_select! (Rust 1.95) without
+# declaring that MSRV. Keep the crate's tested Rust 1.88 floor honest.
+rusqlite = { version = "=0.39.0", features = ["bundled"] }
 nucleo = "0.5"
 clap = { version = "4", features = ["derive"] }
 thiserror = "2"
-anyhow = "1"
+anyhow = "1.0.103"
 dirs = "6"
 chrono = { version = "0.4", features = ["serde"] }
 walkdir = "2"
 rayon = "1"
 unicode-width = "0.2"
+unicode-segmentation = "1.12"
 toml = "1"
 sha2 = "0.10"
 

Cargo.toml

@@ -8,7 +8,7 @@
 > Find the AI coding session you meant to resume.
 
 `agf` is a local-first fuzzy finder for AI coding-agent sessions.
-Search local sessions across **Claude Code**, **Codex**, **Gemini CLI**, **Cursor CLI**, **OpenCode**, **Kiro**, **pi**, and **Hermes** — then resume the right one in a keystroke.
+Search local sessions across **Claude Code**, **Codex**, **Prime Agent**, **Gemini CLI**, **Cursor CLI**, **OpenCode**, **Kiro**, **pi**, and **Hermes** — then resume the right one in a keystroke.
 
 ![agf demo](./assets/demo.gif)
 
@@ -45,10 +45,11 @@ Then you either dig through history files or start over.
 |:---|:---|:---|
 | [Claude Code](https://github.com/anthropics/claude-code) | `claude --resume <id>` | `~/.claude/history.jsonl` + `~/.claude/projects/` |
 | [Codex](https://github.com/openai/codex) | `codex resume <id>` | `~/.codex/sessions/**/*.jsonl` |
+| [Prime Agent](https://github.com/PrimeIntellect-ai/prime-agent) | `prime-agent --resume <id>` | `~/.prime/agent/sessions/<id>.jsonl` |
 | [Gemini CLI](https://github.com/google-gemini/gemini-cli) | `gemini --resume <id>` | `~/.gemini/tmp/<project>/chats/session-*.json` |
 | [Cursor CLI](https://cursor.com/docs/cli/overview) | `cursor-agent --resume <id>` | `~/.cursor/projects/*/agent-transcripts/<id>/<id>.jsonl` (Composer 2+)<br>`~/.cursor/projects/*/agent-transcripts/<id>.txt` (legacy) |
 | [OpenCode](https://github.com/opencode-ai/opencode) | `opencode -s <id>` | `~/.local/share/opencode/opencode.db` |
-| [Kiro](https://kiro.dev) | `kiro-cli chat --resume` *(no per-session resume — always opens the latest session for the cwd)* | `~/Library/Application Support/kiro-cli/data.sqlite3` |
+| [Kiro](https://kiro.dev) | `kiro-cli chat --resume-id <id>` | Kiro v2 SQLite + Kiro v3 `~/.kiro/sessions/cli/` |
 | [pi](https://github.com/badlogic/pi-mono) | `pi --session <id>` | `~/.pi/agent/sessions/<cwd>/*.jsonl` |
 | [Hermes](https://github.com/NousResearch/hermes-agent) | `hermes --resume <id>` *(cwd-independent — resumes in your current shell directory)* | `~/.hermes/state.db` |
 | [Oh My Pi](https://github.com/can1357/oh-my-pi) | `omp --resume <id>` | `~/.omp/agent/sessions/<cwd>/*.jsonl` |
@@ -61,10 +62,11 @@ Then you either dig through history files or start over.
 |:---|:---|:---|
 | Claude Code | JSONL | `~/.claude/history.jsonl` (sessions)<br>`~/.claude/projects/*/` (worktree detection) |
 | Codex | JSONL | `~/.codex/sessions/YYYY/MM/DD/rollout-*.jsonl` |
+| Prime Agent | JSONL | `~/.prime/agent/sessions/<id>.jsonl` (also honors Prime Agent environment/global settings overrides) |
 | OpenCode | SQLite | `~/.local/share/opencode/opencode.db` |
 | pi | JSONL | `~/.pi/agent/sessions/--<encoded-cwd>--/<ts>_<id>.jsonl` |
 | Oh My Pi | JSONL | `~/.omp/agent/sessions/<encoded-cwd>/<ts>_<id>.jsonl` |
-| Kiro | SQLite | macOS: `~/Library/Application Support/kiro-cli/data.sqlite3`<br>Linux: `~/.local/share/kiro-cli/data.sqlite3` |
+| Kiro | SQLite + JSON/JSONL | v2: macOS `~/Library/Application Support/kiro-cli/data.sqlite3`, Linux `~/.local/share/kiro-cli/data.sqlite3`<br>v3: `$KIRO_HOME/sessions/cli/` or `~/.kiro/sessions/cli/` |
 | Cursor CLI | SQLite + JSONL/TXT | `~/.cursor/chats/<workspace>/<id>/store.db` (metadata; required for `.jsonl` to be resumable)<br>`~/.cursor/projects/*/agent-transcripts/<id>/<id>.jsonl` (Composer 2+ transcript)<br>`~/.cursor/projects/*/agent-transcripts/<id>.txt` (legacy transcript) |
 | Gemini | JSON | `~/.gemini/tmp/<project>/chats/session-<date>-<id>.json`<br>`<project>` is a named dir or SHA-256 hash of the project path<br>Project paths resolved via `~/.gemini/projects.json` |
 | Hermes | SQLite | `~/.hermes/state.db` (sessions + messages)<br>JSON dumps in `~/.hermes/sessions/session_<id>.json`<br>Hermes is cwd-independent — resume runs in your current shell directory |
@@ -143,13 +145,14 @@ sort_by = "time"            # "time" | "name" | "agent"
 max_sessions = 200
 search_scope = "name_path"  # "name_path" (default) | "all" (include summaries)
 summary_search_count = 5    # number of summaries included when search_scope = "all"
+include_non_interactive = false # show Codex subagent/exec threads
 ```
 
 You can also edit `search_scope` and `summary_search_count` interactively by pressing `?` in the TUI.
 
 ## Shell integration
 
-`agf setup` auto-detects your shell and installs the wrapper. Supported shells:
+`agf setup` auto-detects your shell and installs the wrapper. Use `agf setup --shell powershell` (or another supported shell) when auto-detection is ambiguous.
 
 - **zsh / bash** — appends to `~/.zshrc` or `~/.bashrc`
 - **fish** — writes to `~/.config/fish/config.fish`
@@ -170,7 +173,7 @@ See [CHANGELOG.md](CHANGELOG.md) for release notes.
 ## Requirements
 
 - macOS, Linux, or Windows (PowerShell 5.1+ / PowerShell 7+)
-- One or more of: `claude`, `codex`, `opencode`, `pi`, `kiro-cli`, `cursor-agent`, `gemini`, `hermes`, `omp`, `yolop`
+- One or more of: `claude`, `codex`, `prime-agent`, `opencode`, `pi`, `kiro-cli`, `cursor-agent`, `gemini`, `hermes`, `omp`, `yolop`
 
 ## Install from source
 
@@ -185,6 +188,8 @@ agf setup
 
 `agf` works best with agents that store resumable sessions locally.
 
+Prime Agent session discovery and resume are supported, but deletion is intentionally disabled: Prime Agent has no public session-delete command and direct file deletion would bypass its active-daemon guard.
+
 **Amp** is not supported yet because its sessions are stored remotely, which makes it hard to reliably resolve local project paths from session metadata. We are monitoring upstream changes and will add support when feasible.
 
 ## Built with

README.md

@@ -40,12 +40,12 @@ Say the new agent is `Foo`, CLI `foo`, sessions under `~/.foo/sessions/`.
    - Bound reads on large transcripts with the shared `read_head_tail` / bounded-read
      helpers in `scanner/mod.rs`; never slurp multi-MB logs whole.
    - Skip malformed lines, don't panic on bad input.
-   - Register the module in **`src/scanner/mod.rs`**: add `pub mod foo;` **and** a
-     `thread::spawn(|| foo::scan().unwrap_or_default())` line in `scan_all()`
-     *(plain `Vec` — the compiler won't remind you)*.
+   - Register the module in **`src/scanner/mod.rs`**: add `pub mod foo;` and an
+     `Agent::Foo => foo::scan()` arm in `scan_agent()`. Do not erase errors into
+     empty success; stale cache rows are retained when a scanner fails.
 
-4. **`src/cache.rs`** — add the `Agent::Foo => scanner::foo::scan().unwrap_or_default()`
-   arm in `start_stale_scan()`. Bump `CACHE_VERSION` only if the cached payload
+4. **`src/cache.rs`** — no agent-specific dispatch is required; workers call the
+   common `scanner::scan_agent()`. Ensure `Session` cache fields round-trip. Bump `CACHE_VERSION` only if the cached payload
    *shape* can change within a single released package version (a new agent key
    alone doesn't require it — the `agf_version` stamp forces a rescan on upgrade).
 
@@ -69,14 +69,14 @@ Say the new agent is `Foo`, CLI `foo`, sessions under `~/.foo/sessions/`.
 
 ```bash
 cargo test --locked
-cargo clippy --locked --all-targets -- -D warnings
+cargo clippy --locked --all-targets --all-features -- -D warnings
 cargo fmt --all -- --check
 # Real-data smoke test (scans all agents regardless of install):
 AGF_DEBUG=1 cargo run -- list --agent foo --format json
 ```
 
-`agf list` runs every scanner unconditionally, so you can verify `foo` against a
-fixture `$HOME` without installing the CLI:
+An explicit `--agent foo` scans only Foo when its cache entry is stale, so you
+can verify it against a fixture `$HOME` without installing the other CLIs:
 
 ```bash
 HOME=/tmp/agf-fixture cargo run -- list --agent foo --format json