diff --git a/script/graphs/tf-main-results.bash b/script/graphs/tf-main-results.bash
new file mode 100644
index 0000000..34ff885
--- /dev/null
+++ b/script/graphs/tf-main-results.bash
@@ -0,0 +1,197 @@
+#!/usr/bin/env bash
+#
+# Copyright (c) 2022 Arm Limited. All rights reserved.
+#
+# SPDX-License-Identifier: BSD-3-Clause
+#
+
+set -euo pipefail
+
+# Build and Jenkins URL.
+sub_build_url=${1}
+job_name="${2:-"tf-a-ci-gateway"}"
+filter=${3:-".*"}
+
+jenkins="${sub_build_url%%/job*}"
+
+# Utilise default paths to output files if none provided.
+job_target="$(dirname ${sub_build_url#*/job/})"
+PNGFILE=${PNGFILE:=${job_target}-result.png}
+CSVFILE=${CSVFILE:=${job_target}-result.csv}
+
+# Remove csv output file if it exists to append to empty file
+: > "${CSVFILE}"
+
+readarray -t sub_builds < <(curl -sSL "${sub_build_url}/api/json" | jq -Rr '
+	fromjson? | [
+		.subBuilds[]? | select(.jobName == "'${job_name}'") | .url
+	] | .[]')
+
+# Get a csv list of relative paths to report.json, or "-" if no report.json
+report_rel_paths_url="${jenkins}/{$(echo $(IFS=,; echo "${sub_builds[*]}"))}/api/json"
+
+readarray -t report_paths < <(curl -fsSL --fail-early "${report_rel_paths_url}" \
+					| sed 's/--_curl_--.*$//' \
+					| sed -e 's/^{/'$(printf "\x1e")'{/' \
+					| jq -sr --seq '
+				 	[ .[]
+						| [ .artifacts[]?
+						| select (.fileName == "report.json")
+						| .relativePath ]
+						| if length > 0 then .[] else "-" end ]
+					| .[]')
+
+# Combine sub build urls with relative path to "report.json"
+# the empty entries are intentionally kept as -, so the output array can
+# be mapped onto ${sub_build_list}
+report_urls="$jenkins/{"
+for i in "${!sub_builds[@]}"
+do
+	if [[ ${report_paths[i]} != "-" ]]
+	then
+		report_urls="${report_urls}${sub_builds[i]}/artifact/${report_paths[i]},"
+	fi
+done
+
+# Strip last comma and add closing brace
+report_urls="${report_urls%?}}"
+
+# Get Child build information from each report.json.
+readarray -t child_file_list_array < <(curl -sL "${report_urls}" -o -\
+                        | sed 's/--_curl_--.*$//' \
+                        | jq -sr --arg FILTER "${filter}" \
+                            '[.[]
+                             | .job as $job
+                             | [ .child_build_numbers?, [(.test_files[]
+                               | sub("\\.test";"")
+                               | split("%") as $config
+                               | { group: $config[1], suite: $config[2]})]]
+                             | transpose
+                             | map( {($job + "/" + .[0]) : .[1]} | to_entries )
+                             | add
+                             | map(select(.value.suite | test(".*"; "il")?
+							               and (endswith("nil")
+										     or endswith("norun-fip.dummy") | not)))
+                             | if ( length > 0 )
+                                then .[] else empty end
+                             | .value.group, (.value.suite | gsub("\\,nil";"")), .key]
+							 | .[]')
+
+# These three arrays should be the same length, and values at the same index
+# correspond to the same child build
+declare -a tftf_keys tftf_suite tftf_group
+
+for i in $(seq 0 3 $((${#child_file_list_array[@]}-1))) ; do
+	tftf_group+=("${child_file_list_array[$i]}")
+	tftf_suite+=("${child_file_list_array[$i+1]}")
+	tftf_keys+=("${child_file_list_array[$i+2]}")
+done
+
+
+child_output_results_url="${jenkins}/job/{$(echo $(IFS=,; echo "${tftf_keys[*]}"))}/api/json"
+
+# Retrieve relative path to either "uart0_full.txt" (FVP) or
+# "job_output.log" (LAVA) for each child job. Once again values where no match
+# is found are intentionally kept as "-" so the array can be correlated with
+# ${tftf_suite}.
+readarray -t child_output_results < <(curl -fsSL --fail-early "$child_output_results_url" \
+                 | sed 's/}{/}\n{/g' \
+                 | jq -sr '[ .[]
+                             | ([ .artifacts[]?
+                                  | select(.fileName == "uart0_full.txt"
+                                  or .fileName == "job_output.log"
+                                  or .fileName == "lava-uart0.log") ]
+                             |  if length > 0
+                                then .[0].relativePath else "-" end), .result ]
+                           | .[]')
+
+# Combine job and child_build number with relative path to output file
+testlog_urls="${jenkins}/job/{"
+tftf_child_results=()
+
+for i in $(seq 0 2 $((${#child_output_results[@]}-1))) ; do
+	testlog_urls+="${tftf_keys[$((i/2))]}/artifact/${child_output_results[$i]},"
+	tftf_child_results+=(${child_output_results[$((i+1))]})
+done
+
+# Remove final comma and append a closing brace
+testlog_urls="${testlog_urls%?}}"
+
+# Retrieve the log for each child with --include to also retrieve the HTTP
+# header and grep for a block like:
+# Tests Skipped : 125
+# Tests Passed  : 45
+# Tests Failed  : 0
+# Tests Crashed : 0
+#
+# If none is found the line HTTP is used to delemit each entry
+#
+# Logs from Lava each message is wrapped with braces and has some pre-amble,
+# which is removed with sed.
+
+tftf_result_keys=(
+        "TestGroup" "TestSuite" "URL" "Result" "Passed" "Failed" "Crashed" "Skipped"
+)
+declare -A results_split="( $(for ord_ in ${tftf_result_keys[@]} ; do echo -n "[$ord_]=\"\" "; done))"
+declare output_csv_str="" csv_row=""
+
+read -ra tftf_urls <<< "$(eval "echo ${testlog_urls}")"
+
+# FIXME adjust this so we can handle both LAVA logs
+# for each test suite
+# curl the result log if its not '-'
+# remove debug information
+# if results is none:
+	# use "Result"
+# else
+	# read each key
+# write row to csv
+
+# Sort results into rows:
+for i in ${!tftf_suite[*]}; do
+	results_split["TestGroup"]="${tftf_group[$i]:-}"
+	results_split["TestSuite"]="\"${tftf_suite[$i]:-}\""
+	results_split["URL"]="${tftf_urls[$i]:-}"
+	results_split["Result"]="${tftf_child_results[$i]:-}"
+
+	# Skipped/Crashed are always zero if no test block is found
+	results_split["Skipped"]="0"
+	results_split["Crashed"]="0"
+	if [[ "${results_split["Result"]}" == "SUCCESS" ]];
+	then
+		results_split["Passed"]="1"
+		results_split["Failed"]="0"
+	else
+		results_split["Passed"]="0"
+		results_split["Failed"]="1"
+	fi
+
+	readarray -t raw_result < <(curl -sL --include "${results_split["URL"]}" \
+			| sed 's/.*msg": "//g' \
+			| grep --text -E "^Tests|HTTP\/")
+
+	for line in "${raw_result[@]}"; do
+		if [[ "${line}" == Test* ]]
+		then
+			k=$(echo "${line}" | awk -F ' ' '{print $2}')
+			count="${line//[!0-9]/}"
+			results_split[$k]=$count
+		fi
+	done
+
+	# Generate CSV row using array of ordinals to align with headers.
+	readarray -t row < <(for k in ${tftf_result_keys[@]} ; do echo "${results_split[$k]}"; done )
+	output_csv_str="${output_csv_str} $(echo $(IFS=,; echo "${row[*]}"))"
+	unset results_split[{..}] row
+done
+
+# Replace spaces in header with commas and print to the output file.
+echo $(IFS=,; echo "${tftf_result_keys[*]}") > ${CSVFILE}
+
+# Sort Filenames alphabetically and store in csv for gnuplot
+sorted=($(IFS=$' '; sort <<<$output_csv_str))
+printf "%b\n" "${sorted[@]}" >> ${CSVFILE}
+
+# Produce PNG image of graph using gnuplot and .plot description file
+gnuplot -e "jenkins_id='$sub_build_url'" -c ${0%bash}plot \
+        "$CSVFILE" > "$PNGFILE"