Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Search across pi session JSONL logs (user prompts, tool calls, results). Uses the session-digest tool and ripgrep for fast triage.
.claude/skills/mikeyobrien-session-search/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | -34% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -53% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -30% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -40% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -42% | 0% |
Search historical pi sessions stored under ~/.pi/agent/sessions/.
This skill assumes the session-digest helper exists at ~/bin/session-digest.
On this Termux setup, sessions are typically stored in:
~/.pi/agent/sessions/--data-data-com.termux-files-home--/*.jsonlIf you are unsure which directory is active, list the session roots:
bashls -la ~/.pi/agent/sessions
session-digest parses session JSONL into readable markdown and writes files to:
~/.rho/digests/YYYY-MM-DD.md1) Generate digests for a time window:
bash# last 7 days session-digest --week >/dev/null # or a single date session-digest 2026-02-04 >/dev/null # or all history (can be slow) session-digest --all >/dev/null
2) Search the digest markdown:
bashrg -n "<query>" ~/.rho/digests
Notes:
session-digest also flags potential secrets.session-digest reports secrets, do NOT paste results into public logs.Search the raw session logs directly:
bashSESSION_ROOT=~/.pi/agent/sessions/--data-data-com.termux-files-home-- rg -n --hidden "<query>" "$SESSION_ROOT"/*.jsonl
Useful variants:
bash# show a little context around matches rg -n -C 2 "<query>" "$SESSION_ROOT"/*.jsonl # case-insensitive rg -n -i "<query>" "$SESSION_ROOT"/*.jsonl
rg -n '"type":"toolCall"' ... | rg '"name":"vault_search"'rg -n '"role":"user"' ... | rg "<phrase>"rg -n "projects/rho" ...session-digest to confirm no secrets are present.| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-03 | fail→fail | 2,692 | 18,902 | +602% | 1 | 1 | 0% | 345 | 839 | +143% | 0 | 0 | — |
case-01 | fail→fail | 3,487 | 10,789 | +209% | 1 | 1 | 0% | 443 | 856 | +93% | 0 | 0 | — |
case-02 | fail→fail | 7,158 | 6,182 | -14% | 1 | 1 | 0% | 632 | 800 | +27% | 0 | 0 | — |
case-04 | fail→pass | 11,132 | 3,722 | -67% | 1 | 1 | 0% | 1,868 | 1,225 | -34% | 0 | 0 | — |
case-05 | fail→pass | 10,701 | 2,374 | -78% | 1 | 1 | 0% | 2,032 | 958 | -53% | 0 | 0 | — |
case-06 | fail→pass | 7,214 | 1,568 | -78% | 1 | 1 | 0% | 1,197 | 841 | -30% | 0 | 0 | — |
case-07 | fail→pass | 8,459 | 1,738 | -79% | 1 | 1 | 0% | 1,387 | 830 | -40% | 0 | 0 | — |
case-08 | fail→pass | 8,040 | 1,896 | -76% | 1 | 1 | 0% | 1,439 | 840 | -42% | 0 | 0 | — |
case-09 | fail→pass | 8,882 | 1,577 | -82% | 1 | 1 | 0% | 1,734 | 814 | -53% | 0 | 0 | — |
case-10 | pass→pass | 4,887 | 2,715 | -44% | 1 | 1 | 0% | 808 | 1,050 | +30% | 0 | 0 | — |
case-11 | pass→pass | 3,328 | 2,636 | -21% | 1 | 1 | 0% | 603 | 1,072 | +78% | 0 | 0 | — |
case-12 | pass→pass | 2,394 | 1,623 | -32% | 1 | 1 | 0% | 368 | 889 | +142% | 0 | 0 | — |
case-13 | pass→pass | 10,840 | 4,679 | -57% | 1 | 1 | 0% | 2,025 | 1,273 | -37% | 0 | 0 | — |
case-14 | fail→pass | 11,274 | 3,060 | -73% | 1 | 1 | 0% | 2,099 | 1,125 | -46% | 0 | 0 | — |
case-15 | fail→pass | 12,339 | 3,816 | -69% | 1 | 1 | 0% | 2,016 | 1,198 | -41% | 0 | 0 | — |
case-16 | pass→pass | 6,984 | 4,461 | -36% | 1 | 1 | 0% | 1,264 | 1,275 | +1% | 0 | 0 | — |
case-17 | fail→pass | 14,062 | 6,798 | -52% | 1 | 1 | 0% | 2,357 | 1,756 | -25% | 0 | 0 | — |
case-18 | pass→pass | 2,965 | 1,445 | -51% | 1 | 1 | 0% | 480 | 801 | +67% | 0 | 0 | — |
case-19 | pass→pass | 8,004 | 6,802 | -15% | 1 | 1 | 0% | 1,345 | 1,749 | +30% | 0 | 0 | — |
case-20 | pass→pass | 3,655 | 2,618 | -28% | 1 | 1 | 0% | 615 | 1,008 | +64% | 0 | 0 | — |
case-21 | pass→pass | 6,749 | 2,996 | -56% | 1 | 1 | 0% | 1,147 | 1,053 | -8% | 0 | 0 | — |
case-22 | fail→pass | 8,224 | 1,248 | -85% | 1 | 1 | 0% | 1,342 | 761 | -43% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted, and 19 counted toward the lift figure. The other 3 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +45 percentage points is the difference between those two pass rates over the 19 comparable cases.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.