profile

Name: profile
Author: Borda

23stars3forks

How to install profile?

borda/ai-rig/profile

$npx -y skills add Borda/AI-Rig --skill profile

Installs into the current project.

Summary

Session clock-time analyzer. Reads the foundry plugin''s timings.jsonl and invocations.jsonl logs (written by task-log.js) and produces a per-session and per-skill wall-time breakdown — local-tool work vs subagent spawns vs Skill invocations vs AskUserQuestion idle vs main-loop r

Files · 1

View on GitHub

SKILL.md

1	<objective>
2
3	Bucket session clock time from existing `~/.claude/logs/{timings,invocations}.jsonl` into:
4
5	1. Local tools — Bash/Read/Edit/Write/Grep/Glob and other main-process tools
6	2. Agent / subagent spawns — Task/Agent calls (sync + background)
7	3. Skill — `tool=Skill` wall durations
8	4. AskUserQuestion idle — human-wait, separate column, excluded from compute total
9	5. Main-loop reasoning (residual) — session wall minus buckets above
10
11	Outputs a markdown report at `.reports/profile/<UTC-timestamp>/report.md` plus a `.temp/output-profile-...md` copy. Includes per-session table, per-skill rollup, and top-N longest single calls.
12
13	NOT for: token or cost accounting (model field null in current logs); per-line Python perf (use `foundry:perf-optimizer`); known failure diagnosis (use `/foundry:investigate`).
14
15	</objective>
16
17	<inputs>
18
19	- `--since DURATION` (default `24h`) — window: `NNs\|NNm\|NNh\|NNd`
20	- `--session-id ID` — optional; restrict to one session
21	- `--top-n N` (default `5`) — slowest single calls to list
22
23	If $ARGUMENTS empty, default window is 24h.
24
25	</inputs>
26
27	<workflow>
28
29	Task tracking: TaskCreate one task ("Run analyzer + render report"); mark in_progress before Bash, completed before final output.
30
31	## Step 1: Parse args + create run dir
32
33	```bash
34	export CSID="${CLAUDE_CODE_SESSION_ID:-$PPID}"
35	# timeout: 5000
36	SINCE="24h"
37	SESSION_ID=""
38	TOP_N="5"
39	for tok in $ARGUMENTS; do
40	case "$tok" in
41	--since=*) SINCE="${tok#--since=}" ;;
42	--since) next_is_since=1 ;;
43	--session-id=*) SESSION_ID="${tok#--session-id=}" ;;
44	--top-n=*) TOP_N="${tok#--top-n=}" ;;
45	*)
46	if [ "${next_is_since:-0}" = "1" ]; then SINCE="$tok"; next_is_since=0; fi
47	;;
48	esac
49	done
50	STAMP="$(date -u +%Y-%m-%dT%H-%M-%SZ)"
51	REPORT_DIR=".reports/profile/$STAMP"
52	mkdir -p "$REPORT_DIR"
53	{
54	echo "REPORT_DIR=$REPORT_DIR"
55	echo "SINCE=$SINCE"
56	echo "SESSION_ID=$SESSION_ID"
57	echo "TOP_N=$TOP_N"
58	} \| tee "${TMPDIR:-/tmp}/foundry-profile-state-${CSID}"
59	```
60
61	Values persisted to `${TMPDIR:-/tmp}/foundry-profile-state-${CSID}`; Steps 2–3 re-source it (bash state does not persist across Bash calls, and `REPORT_DIR` carries a per-shell timestamp that cannot be re-derived).
62
63	## Step 2: Run analyzer
64
65	```bash
66	export CSID="${CLAUDE_CODE_SESSION_ID:-$PPID}"
67	# timeout: 60000
68	. "${TMPDIR:-/tmp}/foundry-profile-state-${CSID}" 2>/dev/null # reload REPORT_DIR/SINCE/SESSION_ID/TOP_N (fresh shell)
69	OPT_SID=""
70	[ -n "$SESSION_ID" ] && OPT_SID="--session-id $SESSION_ID"
71	python "${CLAUDE_PLUGIN_ROOT:-plugins/cc_foundry}/bin/timing_analyzer.py" \
72	--since "$SINCE" \
73	--top-n "$TOP_N" \
74	--output "$REPORT_DIR/report.md" \
75	$OPT_SID 2>"$REPORT_DIR/warnings.log"
76	echo "exit=$?"
77	```
78
79	`OPT_SID` left unquoted so empty expands to nothing (no flag). Exit code 1 → no sessions in window — surface that and stop.
80
81	## Step 3: Mark run complete
82
83	```bash
84	export CSID="${CLAUDE_CODE_SESSION_ID:-$PPID}"
85	# timeout: 5000
86	. "${TMPDIR:-/tmp}/foundry-profile-state-${CSID}" 2>/dev/null # reload REPORT_DIR/SINCE/SESSION_ID/TOP_N (fresh shell)
87	echo '{"status":"complete","since":"'"$SINCE"'","session_id":"'"$SESSION_ID"'","top_n":'"$TOP_N"'}' > "$REPORT_DIR/result.jsonl"
88	```
89
90	## Step 4: Emit terminal output
91
92	Read YAML header from `$REPORT_DIR/report.md` (first block between `---` lines) and print verbatim. Then print `→ $REPORT_DIR/report.md`. Then read Headline split block plus top 3 sessions from per-session table and surface as executive summary (per quality-gates.md output routing).
93
94	Also Write the long-output dump per quality-gates rule:
95
96	```text
97	Write(file_path=".temp/output-profile-<branch>-<YYYY-MM-DD>.md", content=<full report contents>)
98	```
99
100	Where `<branch>` = `$(git branch --show-current 2>/dev/null \| tr '/' '-' \|\| echo 'main')`.
101
102	## Step 5: Follow-up gate
103
104	Invoke `AskUserQuestion`:
105	- (a) Drill into slowest session — re-run

People who liked this also viewed

6 picks

Type

profile

23stars3forks

View on GitHub

How to install profile?

borda/ai-rig/profile

$npx -y skills add Borda/AI-Rig --skill profile

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	<objective>
2
3	Bucket session clock time from existing `~/.claude/logs/{timings,invocations}.jsonl` into:
4
5	1. Local tools — Bash/Read/Edit/Write/Grep/Glob and other main-process tools
6	2. Agent / subagent spawns — Task/Agent calls (sync + background)
7	3. Skill — `tool=Skill` wall durations
8	4. AskUserQuestion idle — human-wait, separate column, excluded from compute total
9	5. Main-loop reasoning (residual) — session wall minus buckets above
10
11	Outputs a markdown report at `.reports/profile/<UTC-timestamp>/report.md` plus a `.temp/output-profile-...md` copy. Includes per-session table, per-skill rollup, and top-N longest single calls.
12
13	NOT for: token or cost accounting (model field null in current logs); per-line Python perf (use `foundry:perf-optimizer`); known failure diagnosis (use `/foundry:investigate`).
14
15	</objective>
16
17	<inputs>
18
19	- `--since DURATION` (default `24h`) — window: `NNs\|NNm\|NNh\|NNd`
20	- `--session-id ID` — optional; restrict to one session
21	- `--top-n N` (default `5`) — slowest single calls to list
22
23	If $ARGUMENTS empty, default window is 24h.
24
25	</inputs>
26
27	<workflow>
28
29	Task tracking: TaskCreate one task ("Run analyzer + render report"); mark in_progress before Bash, completed before final output.
30
31	## Step 1: Parse args + create run dir
32
33	```bash
34	export CSID="${CLAUDE_CODE_SESSION_ID:-$PPID}"
35	# timeout: 5000
36	SINCE="24h"
37	SESSION_ID=""
38	TOP_N="5"
39	for tok in $ARGUMENTS; do
40	case "$tok" in
41	--since=*) SINCE="${tok#--since=}" ;;
42	--since) next_is_since=1 ;;
43	--session-id=*) SESSION_ID="${tok#--session-id=}" ;;
44	--top-n=*) TOP_N="${tok#--top-n=}" ;;
45	*)
46	if [ "${next_is_since:-0}" = "1" ]; then SINCE="$tok"; next_is_since=0; fi
47	;;
48	esac
49	done
50	STAMP="$(date -u +%Y-%m-%dT%H-%M-%SZ)"
51	REPORT_DIR=".reports/profile/$STAMP"
52	mkdir -p "$REPORT_DIR"
53	{
54	echo "REPORT_DIR=$REPORT_DIR"
55	echo "SINCE=$SINCE"
56	echo "SESSION_ID=$SESSION_ID"
57	echo "TOP_N=$TOP_N"
58	} \| tee "${TMPDIR:-/tmp}/foundry-profile-state-${CSID}"
59	```
60
61	Values persisted to `${TMPDIR:-/tmp}/foundry-profile-state-${CSID}`; Steps 2–3 re-source it (bash state does not persist across Bash calls, and `REPORT_DIR` carries a per-shell timestamp that cannot be re-derived).
62
63	## Step 2: Run analyzer
64
65	```bash
66	export CSID="${CLAUDE_CODE_SESSION_ID:-$PPID}"
67	# timeout: 60000
68	. "${TMPDIR:-/tmp}/foundry-profile-state-${CSID}" 2>/dev/null # reload REPORT_DIR/SINCE/SESSION_ID/TOP_N (fresh shell)
69	OPT_SID=""
70	[ -n "$SESSION_ID" ] && OPT_SID="--session-id $SESSION_ID"
71	python "${CLAUDE_PLUGIN_ROOT:-plugins/cc_foundry}/bin/timing_analyzer.py" \
72	--since "$SINCE" \
73	--top-n "$TOP_N" \
74	--output "$REPORT_DIR/report.md" \
75	$OPT_SID 2>"$REPORT_DIR/warnings.log"
76	echo "exit=$?"
77	```
78
79	`OPT_SID` left unquoted so empty expands to nothing (no flag). Exit code 1 → no sessions in window — surface that and stop.
80
81	## Step 3: Mark run complete
82
83	```bash
84	export CSID="${CLAUDE_CODE_SESSION_ID:-$PPID}"
85	# timeout: 5000
86	. "${TMPDIR:-/tmp}/foundry-profile-state-${CSID}" 2>/dev/null # reload REPORT_DIR/SINCE/SESSION_ID/TOP_N (fresh shell)
87	echo '{"status":"complete","since":"'"$SINCE"'","session_id":"'"$SESSION_ID"'","top_n":'"$TOP_N"'}' > "$REPORT_DIR/result.jsonl"
88	```
89
90	## Step 4: Emit terminal output
91
92	Read YAML header from `$REPORT_DIR/report.md` (first block between `---` lines) and print verbatim. Then print `→ $REPORT_DIR/report.md`. Then read Headline split block plus top 3 sessions from per-session table and surface as executive summary (per quality-gates.md output routing).
93
94	Also Write the long-output dump per quality-gates rule:
95
96	```text
97	Write(file_path=".temp/output-profile-<branch>-<YYYY-MM-DD>.md", content=<full report contents>)
98	```
99
100	Where `<branch>` = `$(git branch --show-current 2>/dev/null \| tr '/' '-' \|\| echo 'main')`.
101
102	## Step 5: Follow-up gate
103
104	Invoke `AskUserQuestion`:
105	- (a) Drill into slowest session — re-run

People who liked this also viewed

6 picks

Type